This PDF file contains the front matter associated with SPIE Proceedings Volume 12257, including the Title Page, Copyright information, Table of Contents and Conference Committee lists.

Currently, depth estimation of 2D image is widely treated as an important technology for environmental perception in autonomous driving, but it still suffers from many issues. From the view of application, this work proposes an unsupervised monocular depth estimation based on hybrid ViT to improve accuracy and reduce cost. Specifically, the technology of convolution and transformer have been combined in this work to encode to extract fine-grained features. Besides, fusing multi-scale features to decode is also adopted to generate multi-scale disparity maps. Then, the loss is calculated based on multi-scale and full-resolution disparity maps, and stereo constraints to realize image reconstruction have been achieved finally. Additionally, experiments have been carried out on the KITTI dataset, and the measured results indicate that compared with the previous works, this work has made progresses in the indicators of error and accuracy, i.e., higher accuracy of 3.4% than baseline, more clear boundaries, fewer artifacts and higher quality of depth maps. It is proved that the combination of hybrid encoding, multi-scale decoder and full-resolution loss can bring significant effect on depth estimation, especially the hybrid encoding.

As a decision-making layer, path planning technology plays an important role in intelligent driving. Using the A* algorithm or the TEB (timed elastic band) algorithm alone cannot achieve global path optimization and dynamic obstacle avoidance at the same time. This paper adopts a path planning strategy combining A* and TEB algorithms. First, use the A* algorithm to test in a static environment, then use the TEB algorithm to test in a dynamic environment, and finally, conduct experiments in a complex environment with both static and dynamic obstacles, the feasibility and effectiveness of the method for path planning in static maps with dynamic obstacles are verified.

In this paper, genetic algorithm is used to improve LSTM neural network, and then a photovoltaic power prediction model is obtained. After preprocessing and normalizing the original power generation, The LSTM considering both time and nonlinearity is used for training, and finally it is improved by genetic algorithm. The optimized LSTM model is applied to the problems in short-term prediction of photovoltaic power generation, and compared with traditional methods, so as to achieve more accurate prediction results.

In the real applications, labeled data is often in short supply, which make is difficult to train a reliable classifier. Transfer learning theory enables model to be trained by exploiting the auxiliary data from out domains. For transfer learning, it’s crucial to find a new feature representation where traditional classifier can be applied. In this paper, we propose a novel transfer algorithm by constructing a Laplacian feature embedding where data property is preserved and domain discrepancy is reduced by MMD principle. Furthermore, we exploit the label constraint during the embedding procedure which minimizes the empirical error on the labeled data. Extensive experiments are conducted to evaluate the effectiveness on two textual datasets, and the study shows the proposed method can significantly improve the classification accuracy over several state-of-the-art algorithms.

The road condition in the field environment is complex, and the accurate localization of unmanned vehicle is critical to solve the SLAM (Simultaneous Localization and Mapping) problem. In this paper, GPS, IMU and odometer sensors are selected as hardware equipment, and the strategy of combining global positioning and local positioning is formulated. The sensor data fusion is utilized with the more adaptive unscented Kalman filter algorithm as the core. The vehicle experiment is carried out in the real field environment to ensure that the localization system can precisely output the coordinates of unmanned vehicles under the condition of high nonlinearity as well as meet the robustness and accuracy requirements of unmanned vehicle localization in the field environment.

Considering the information of video data spatially and temporally, this paper proposes a novel method of simple action recognition based on spatio-temporal features. For each frame, the spatial feature sequence is built by the joint angle features and the joint distance features after the human skeleton is obtained by the lightweight OpenPose. The atomic action is classified via the spatial feature sequences from video frames. Then, the atomic action label sequences are used to train hidden Markov models (HMMs) so that the constructed models can suit each action. This approach presents the advantages of the full use of the spatial features and the excellent learning ability of HMM. Experiments on datasets demonstrate the accuracy in simple action recognition of the proposed method.

Objective: To review the application of VR technology in anxiety disorders for reference in the industry.Methods: This paper searched the relevant research documents at home and abroad from 2011 to 2021, and analyzes the changes of the number of virtual reality technology in the number of anxiety disorder research, the application of virtual reality technology in the treatment of anxiety disorder, and the distribution group of the research.Results: A total of 717 documents were retrieved, mainly involving three groups: anxiety patients, different types of clinical patients, and healthy people with social anxiety. The use of virtual technology mainly included the separate use of virtual technology and intervention research combined with traditional methods.Conclusion: Virtual technology has good applicability in different types of anxiety groups, and its intervention effect is more significant than conventional methods, moreover VR combining with conventional training methods can get better effect as well .Virtual technology has broad application prospects. With the maturity of this technology, the virtual reality technology can be combined with more different treatment methods to further apply and study, and continuously expand the scope of the research group.

The rapid progress of Information Technology has brought tremendous technological changes in the language service sector. Terminology management with computer-aided translation support is a "powerful tool" for translators to improve translation efficiency and ensure translation quality. Rational and systematic terminology management can enhance accuracy, consistency, and sustainable use of translation terms and ensure readable and scientific translation. This paper has elaborated on terminology management and its significance in translation. Taking the Trados as an example, it has analyzed and discussed the terminology management application and its positive effects on improving translation quality.

With the continuous advancement of transmission data rate, high-speed electronic interconnections are extensively investigated in electronic and information engineering field. For a high-speed data transmission system, channel degradation may result in deterioration of communication quality of digital signals. In X-band and Ku-band, high-speed interconnected transmission systems using single sideband modulation were proposed in this paper, where Pseudo Random Binary Sequence data with 231-1 pattern was used as the signal source at the rate of theoretical maximum. With regard to the comparison between pristine and degraded channels, different Ball Grid Arrays high-frequency transmission channels were respectively fabricated and measured. Experimental measurements of channels demonstrate that failed solder balls would cause significant decrease of reflection and transmission parameters in Ku-band. Simulations of all system models indicate that excellent eye openings and low bit error rates could be obtained, although the eye height was deteriorated in those systems working in Ku-band. Based on experimental tests and simulations, the results of this investigation provide a better understanding and evaluating for the effects of degraded interconnection channel in high-speed data transmission systems.

Automated segmentation of liver and lesion in contrast-enhanced abdominal computed tomography (CT) scans is a crucial step for computer-assisted diagnosis and surgical planning of liver diseases. Although deep convolutional neural networks (DCNNs) have contributed many breakthroughs in liver and lesion segmentation task from CT images, the problem remains challenging due to low segmentation accuracy and high complexity. To address these issues, we proposed a light-weight convolutional neural networks for liver and lesion segmentation from CT images, i.e., LW-Mnet. Firstly, we improve the basic module of MobileNetv3 as the backbone of LW-Mnet. The second is that we employ LWASPP to capture multi-scale feature information for enhance representation power of network. The proposed network is evaluated on the public LiTS dataset, and the results show that LW-Mnet achieved the better performance than other networks with fewer parameters and faster inference speed.

Virtual classroom is a challenging research topic which is often discussed. It may be one part of metaverse. In this paper, we introduce 3D shape estimation for the registration and authentication of teachers and students in a virtual classroom, a new deep learning method, called geometric generative model, and 2D to 3D mapping algorithm DensePose are used. In the Login of the virtual classroom, the 3D human body reconstructed by the 2D image input of the teacher or the student is used as the message base of the identification and the authentication. Experiments of the first step of the virtual classroom, that is login and authentication based on 3D Shape model of the teacher and student is going to be done.

Bilinear networks generally only do bilinear pooling operation on the feature vector after the last layer is processed by the convolutional layer of the activation function, and use the fused features for fine-grained image classification. However, this approach ignores the feature information of the middle convolutional layer, which cannot completely describe each semantic part of the image and easily causes the loss of discriminative information of fine-grained categories. In this paper, we propose a fine-grained image classification method based on multilayer feature fusion, in which the ResNet model is selected to replace the two-way VGGNet model in the original bilinear convolutional neural network to improve the feature extraction ability of the network, and the different layer feature outputs extracted from the last three convolutional residual blocks of ResNet are bilinearly pooled, and the three feature vectors are stitched together for multilayer The resulting three feature vectors are stitched together for multilayer feature fusion and then fed to the classifier for classification.

In order to objectively and quantitatively evaluate the facial morphology of patients in plastic surgery, the key points of facial aesthetics are extracted by interaction and specified face database, and the geometric features are calculated to obtain accurate quantitative face data, and the aesthetic evaluation model is established. Face assessment, eyebrows and eyes overall assessment and nose assessment were carried out for the subjects. The results showed that the model could objectively give the facial score of the subjects, find out the deviation from the standard, and provide objective and effective guidance for further plastic surgery.

Using wireless sensor network, embedded, cellular network and data acquisition technologies, with CC2530 low-power microcontroller chip as the control unit, with WinCE real-time multitasking operating system as the management platform, through ZigBee wireless network to collect and transmit terminal data, contains temperature and humidity data, carbon dioxide concentration, soil nitrogen, phosphorus, and potassium concentration, etc. The PHPRPC protocol is used for communication and data exchange between software, and remote monitoring based on the Web is realized. Achieve higher real-time, security and stability in agricultural monitoring.

The fault diagnosis of complex system is of great significance. In view of the current problems such as little fault data of newly researched equipment. Taking the engine drive pump, the core component in the hydraulic system, for example, a digital model of fault conditions is established and fault data is generated through batch simulation, and multiple types of fault classifiers such as leakage, blockage and rotating shaft wear are developed by extracting the state index of pump output flow, so as to detect different faults. The test results show that the classifier obtained by model simulation of healthy fault data can detect faults more accurately, which solves the problem of difficult fault diagnosis caused by the lack of test data of key components of hydraulic system and effectively improves the level of fault diagnosis and maintenance of aviation equipment.

The ROAD meta-architecture guides the whole enterprise architecture development process with business architecture. In response to the problem that the IDEF0 used in the business activity model has insufficient ability to describe the rich activity trigger sequences and information interaction scenarios between multi-component systems, this paper introduces the scenario model, which is fused with the original IDEF0-based activity model to become a scenario-oriented activity model with hierarchical description capability and scenario timing description capability. The extended ROADS metaarchitecture enables scenario-oriented system modeling and realizes scenario management in business architecture design. Finally, the kindergarten health system is used as an example for validation.

Diabetic foot is a chronic diabetic complication. Infrared thermography can be used to scan the bottom of the foot to obtain a thermographic image, and identifying the location of both feet in the image is a key step. A deep learning based algorithm is proposed for automatic recognition of the obtained thermograms. The plantar thermograms of different patients are made into a dataset, trained using deep learning algorithms, and the trained model is judged by the obtained loss function profile, and then analyzed for bipedal recognition. Finally, the experiments compare the recognition performance of Mask-RCNN and Hybrid-TaskCascade (HTC) deep learning algorithms, and the results show that the former has higher detection speed and accuracy. The recognition rate of the former is 99.96%, which meets the require- ment of high recognition rate of biped in thermograms.

Concealment camouflage efficacy is the focus of system operation test, System of systems operation and single equipment operation test. In recent years, new equivalent reconnaissance means have emerged, whether the equivalent effect is consistent with the actual combat effect can’t be evaluated. The evaluation of concealment camouflage effect has not been solved. It is urgent to create a new evaluation model to support the test of concealment camouflage effect. Aiming at the new equivalent reconnaissance technology, this paper studies and establishes the equivalent effect evaluation model and concealment camouflage capability evaluation model, which makes an effective exploration and attempt to solve the current problem of concealment camouflage efficacy evaluation.

With a view to the accuracy of identifying ship hydrodynamic parameters using least squares support vector regression (LS-SVR), a LS-SVR based on multi-excitation difference method is proposed. In this method, the aperiodic excitation samples composed of zigzag maneuvering and rotary maneuvering are used as the input samples. Meanwhile, considering the parameter drift caused by the multicollinearity of the identified samples in the ship hydrodynamic model, the difference method is used to reduce the correlation coefficient between the input samples. The simulation study on the Abkowitz dynamics model of Mariner is carried out to test the proposed LS-SVR based on multi-excitation difference method. The simulation results show that the proposed identification method which is LS-SVR based on multi excitation difference method is more accurate through the comparison with LS-SVR base on zigzag maneuvering input samples using single periodic signal.

Inflow is an important factor affecting the accuracy of power generation dispatching of hydropower stations. Due to the randomness of Inflow, there are usually deviations in the runoff prediction of hydropower stations, especially for hydropower stations in large watersheds, the deviation of runoff prediction may greatly affect power generation dispatching. In this paper, a deviation analysis method of runoff prediction of hydropower stations in large watershed under the condition of random inflow is proposed. The ARIMA model in the time series analysis method is used for runoff prediction, and then the error distribution law between the prediction results and the actual results is analyzed by copula function, and the error distribution of annual and monthly runoff prediction and power generation capacity prediction of cascade hydropower stations is analyzed. Taking Xiluodu, Xiaowan and Nuozhadu hydropower stations as examples, the annual runoff prediction and error distribution law are analyzed, which can accurately identify the range of deviation.

With the rapid development of UAVs, the application of UAVs in civil and military fields has greatly been expanded. UAVs are often used to perform tasks in hazardous areas because of its flexibility. Although many different path planning algorithms have been proposed to find optimal or near-optimal collision-free paths for UAVs, most of them either do not consider dynamic obstacle avoidance or ignore nonlinear kinematic characteristics of UAVs, which is impractical. In this paper, a Deep Reinforcement Learning (DRL) based path planning algorithm with the nonlinear kinematic constraints is proposed. For static obstacles and task points, the reward function is designed and DRL is used to generate the flight path for multiple task points. When UAV detects a dynamic obstacle, we assume that it transmits coordinates of the obstacle to a computer center to regenerate the remaining path. Wilson Confidence Interval (WCI) is then used to predict time taken by computer center to recreate the new path. Simulation results validate the efficiency of the proposed algorithm.

The paper discusses an ultra-short-term wind power prediction model that optimizes the parameters of the LSTM model by improving the whale algorithm. After the initial wind power-related data is pre-processed and normalized, the training model is constructed by using the long-term short-term neural network, LSTM neural network, whale algorithm optimization, ultra-short-term wind power account for optimization LSTM. The results show that the difference is positive, so its correctness, effectiveness, and optimization parameters can be considered.

Global navigation satellite system (GNSS) signals are easily blocked and reflected by surroundings in urban environments, resulting in a large number of non-line-of-sight (NLOS) signals and multipath effects, which will extremely degrade GNSS positioning precision and reliability. This paper proposes the GNSS NLOS signals identification method based on the fully connected neural networks (FCNNs) in deep neural networks (DNN), which extracts representative features from GNSS original observations, and uses the DNN to realize effectively signal classification tasks. This method improves the precision and reliability of GNSS positioning by predicting the NLOS signals accurately. The experimental results show that the Doppler measurement, C/N0, elevation angle, azimuth, and the NLOS visibility ratio can be used as representative parameters of the current NLOS identification task. Compared with the decision tree, and the support vector machine, the experimental results show that the prediction effect of the FCNNs algorithm is better than the former two algorithms.

Aiming at the automatic reading task of pointer meter in industrial applications, in order to solve the problems of indistinct meter features, complex background environment and slow recognition speed involved in the reading process of pointer meter, a new type of U-net network is proposed to segment Instrument hands and scales. The network replaces the backbone feature extraction part in the traditional U-net network with the bneck module in MobileNet, and fuses it with the enhanced feature extraction part. First, use web crawlers and data enhancement to expand the dataset; second, use the YOLO network to filter out the complex background information of the collected images and extract the dashboard area. Input the dial area information into the improved version of the U-net network to segment the meter scale and pointer; finally, convert the annular image into a rectangular image, and obtain a one-dimensional array of scale and pointer values through the rectangular image, and then calculate Pointer meter readings. Theoretical analysis and experimental results show that the new U-net can segment the pointer and scale of the meter more accurately, and the segmentation accuracy is increased from 90.79% of the traditional U-net network to 93.63%. The proposed automatic reading method of the pointer meter has good performance. The accuracy and anti-interference ability of the system, the average basic error is 0.48%, which meets the requirements of the application.

To solve the control problem of formation with limited information of mobile robots, a formation control method based on vision and azimuth measurement information is proposed. In this method, the relative position between the leader and the follower is estimated by using the coordinates of the landmarks of the leading robot on the image plane of the onboard camera of the follower, and the azimuth information of the robot is obtained by using the on-board sensor of the robot. The leader-following formation controller is designed according to the relative position estimation and the measuring azimuth information. Lyapunov stability theory is used to prove the system stability. The joint simulation results of Coppeliasim and Matlab show that formation control method based on vision and azimuth measurement information can make multiple mobile robots form stable formation quickly.

The Chinese Electronic Medical Records (EMR) lacks abundant annotated data. This severely limits the performance of Named Entity Recognition (NER) models in this domain. We propose a Chinese electronic medical record named entity recognition model based on pre-training and multi-task learning (Pt-Mt) to solve this problem. The model first fine-tunes the improved pre-trained model Roberta on different medical datasets, so that Roberta better fits the characteristics of the medical field and can learn features on different datasets. At the same time, the Chinese Word Segmentation (CWS) task is added as an auxiliary task of the NER model for joint training, which enhances the model's ability to distinguish entity boundaries. Finally, the NER of Chinese EMR based on Roberta and multi-task learning has achieved good results on the CCKS2017, CCKS2019, and CCKS2020.

Data quality is very important in data centric environments. The data quality profiling task is to filter out data records which violate domain semantic rules. The data quality profiling framework is expected to be flexible to represent complex domain semantics and efficient to tackle large amount of data records. We propose a data quality profiling framework, named MyETL, which is conceived from ETL paradigm to fulfil these requirements. A directed acyclic graph is employed to represent domain semantic rules in design phase. Then the graph is optimized by a topology optimization procedure. At last, the graph is mapped to threads and memory objects and scheduled to execution. As implemented based on OSGi framework, MyETL is constructed by bundles and can be extended for convenient.

In this paper, a novel intelligent monitoring system based on digital twin framework is developed for spacecraft vacuum thermal test. More precisely, the corresponding digital twin model mapping of real-world spacecraft vacuum thermal test is constructed by hierarchy design scheme with robustness. Then, in order to reliably obtain all the dynamical data during the entire vacuum thermal testing processes, the monitoring system is designed with hardware equipment and software integration. As a result, the testing data analysis, fault detection and decision optimization for spacecraft vacuum thermal test can be achieved, which can effectively save the implementation cost and improve the testing efficiency. In the end, the practical experiment case is presented for showing the usefulness of the designed intelligent monitoring system.

Time domain broadband beamforming algorithm can improve the processing performance of sonar array system because it can output continuous waveforms in time domain. In view of the narrow null suppression in the interference direction and the shallow null depth limited by the interference power in traditional beamforming, an algorithm based on null broadening and null deepening is proposed to improve the robustness of the beam in the interference direction, and the time-domain waveform distortion suppression of the beamformer is increased by reducing the main lobe width, The simulation results show its effectiveness.

Under the environment of energy crisis and green development, the development of electric vehicles has become inevitable. However, the charging load of electric vehicles has great uncertainty, which will have a certain impact on the security and stability of the power grid. Accurately predicting the charging load of electric vehicles is an effective method to avoid this problem. Considering the spatiotemporal characteristics of electric vehicle charging load, an electric vehicle load prediction method based on differential evolution algorithm improved BP neural network is proposed to complete the search of the weight space and network structure space of the neural network at the same time. The optimal network structure. The algorithm adopts the (1+1)-ES binary evolution strategy, uses a new network structure crossover and mutation method, and speeds up the search of the neural network model and the algorithm through the co-evolution of dual population structure and adaptive mutation rate strategies. Convergence improves the learning ability of the network and reduces the prediction error of the BP neural network model. The model and the BP neural network model are respectively used to predict the load of electric vehicles, and the comparison results prove the superiority of the proposed model.

Image processing is a key link in the process of automobile intelligence. Image pretreatment methods are studied, including gray processing, binarization and extraction of regions of interest. Three different methods of gray image are compared, and the maximum interclass variance method is used for binarization processing of images. When the image is too bright or too dark, the recognition is very inaccurate. Then the edge detection and straight line acquisition were studied to identify the lane lines, and the Canny edge detection was studied, including Gaussian filtering smoothing, Sobel operator, non-maximum suppression and double threshold edge extraction, and the Canny algorithm was improved and optimized to study the Hough transform. Hough transform is used to extract the straight lines to recognize the lane lines.

The terminal sliding mode control can effectively eliminate the uncertainties such as the parameter perturbation and the external load disturbance of a direct current (DC) torque motor, but still exists the chattering and the slow convergence speed, which affect the stability and speed of motor tracking control system. Therefore, a fast nonsingular terminal sliding mode (FNTSM) control strategy based on improved power reaching law is proposed. A fast convergent exponential function is introduced into the double power reaching law, which makes the whole control strategy not only improves the convergence speed when the system reaches the sliding surface but also inherits the ability of power term to suppress chattering. The theoretical analysis and comparative simulation prove that, in the case of load disturbance, the FNTSM control method using the improved power reaching law not only shortens the convergence time and has faster convergence speed, but also suppresses the chattering of the system, so that the DC motor can track the given trajectory quickly and accurately. Therefore, it’s confirmed that the proposed control method owns more effectiveness and superiority.

In the context of construction engineering, the classification and calculation of the amount of decoration has become the main problem. The structure of engineering drawings is complex and the drawing standards are not unified. This paper proposed a structure extraction algorithm to filter interference lines, extract wall structures and divide spatial regions through image morphology processing based on the image features of floor plan. Based on the target detection algorithm, the drawing scale detection and recognition are realized, and the spatial area calculation is completed through the coordinate system transformation. The results are evaluated by the evaluation standard of graph recognition performance and annotation verification set. The algorithm proposed in this paper is superior to other algorithms in spatial structure analysis, and can effectively analyze and calculate the spatial structure of floor plan in architectural engineering scenes.

Bayesian knowledge tracking model is used to track learners' knowledge state and predict their mastery level and future performance in intelligent teaching system. The original BKT model assumes that learners do not forget any knowledge after learning. This assumption will lead to the deviation between the predicted results of the model and the actual situation. In order to deal with above situations, this paper proposes a Bayesian knowledge tracking model based on learner behavior and forgetting factors. By using the decision tree algorithm to obtain the behavior node data information, and then initialize the forgetting parameters and assign values to update the algorithm of learners' knowledge mastery level.

Cross-modal retrieval refers to accurate retrieval between data from different modes. Its core task is how to measure the similarity between different modal data across semantic divides. In this paper, a novel cross-modal retrieval model CLCMR is proposed, which makes the identification of similar samples of different images more inclusive and the identification of different samples more discriminative. Specifically, the comparative learning method is introduced into the model, the pretraining dataset is processed by the data enhancement method, and the image Encode is pre-trained to make the image features extracted by the same sample closer The difference in image characteristics of similar samples improves their recognition accuracy and accuracy. In the downstream task, in order to ensure the effect of Encode, three loss functions are selected to ensure the nativeity of the data mapping of each mode into the common space, that is, to reduce the modal loss caused by the transformation space. Finally, in order to semantically align images and text, the content of Fast-RCNN was added to the CLCMR network framework for cross-modal retrieval. In this paper, several experiments are conducted on the Pascal Sentence dataset and the XmediaNet dataset, and the results show that the CLCMR cross-modal retrieval framework proposed in this paper is used The mAP value is better than some of the methods commonly used at present, which verifies the feasibility and perfection of the method.

Recently, the attention mechanism and convolutional operation have been widely applied in Chinese Named Entity Recognition (NER) owing to their parallelization ability. However, in Chinese NER, the attention mechanism tends to model for global information and almost ignores local feature information between characters. On the contrary, convolutional operation can capture the local information while inability to solve patterns with discontinuous characters. In this paper, we propose a novel dual stream feature fusion encoding method. Specifically, we design a Dual-stream Temporal Network (DSTN), in which we use the advantages of both convolutional operation and self-attention mechanism while alleviate their respective drawbacks. DSTN can effectively capture the local and global feature information by encoding the characters. Besides, we also present a loss calculation method, namely Multi-loss, which can prevent the model from over-fitting. The experiment results on two NER datasets showed that our method has excellent performance and efficiency than most mainstream methods.

The Continuous-variable quantum key distribution (CVQKD) technique is one of the practical uses of quantum information technology. CVQKD system is implemented based on existing optical networks and standard telecommunication technologies. CVQKD can be put into use of the information transmission in city-based areas, within the extent of absolute security and without any restrictions of security. To reduce the experimental system to a commercial prototype, a data acquisition (DAQ) card can be used to replace the DAQ equipment in the laboratory. It can save the cost of the system experiment and commercial application. In this paper, through the high-speed DAQ card, we implement the collection and uploading of the electrical signal, when the light passes through the continuous optical coherent state detector. The whole experiment test system consists of a field-programmable gate array (FPGA), an analog-to-digital converter (ADC) chip, and optical components. This paper provides a method of gain adjustment of the balanced homodyne detector and a way to test the π phase voltage of the optical modulator. The results show that the DAQ card can be used in the development of the CVQKD prototype.

With the rise of virtual reality technology and eye movement tracking, the cognitive classification of eye movement emotion combined with virtual reality technology was a research hotspot recently. In this research, 96 pictures were selected from the Chinese Affective Picture System (CAPS), and 35 subjects were stimulated with pictures in virtual scene. The corresponding eye movement data were tracked synchronously. Eye movement data were progressed with blink data reconstruction and baseline correction, after which time and frequency domain features were extracted. Finally, for three emotion model dimensions including valence, arousal and dominance expressed as high, medium and low, models are trained with Support Vector Machine (SVM), with the highest classification accuracy of single person reaches 78.1%. The results of this paper are instructive for the processing of eye movement data, and the eye tracking data combined with virtual reality technology can be used for emotion recognition research through the proposed preprocessing methods and feature extraction.

In this paper, the 50MVA/110kV oil-immersed transformer is taken as the research object. Firstly, the three-dimensional model of the transformer is established according to the actual size, and the corresponding boundary conditions are set according to the relevant theory of electromagnetic field. The finite element method is used to analyze the internal electrical quantity and magnetic field distribution of the transformer during operation. According to the theory of heat transfer and fluid mechanics, considering the nonlinear characteristics of transformer oil with temperature changes, a simulation model of transformer heat-flow is built, The transformer loss is introduced into the simulation model as a heat source, and then the transformer electromagnetic-thermal-fluid coupling is realized. Combined with machine learning algorithm and multi-physics coupling calculation, the data obtained from transformer simulation and actual optical fiber temperature measurement are combined to form a data sample set, taking the load rate, ambient temperature, and top oil temperature as the input feature quantities, and the hot spot temperature as the output quantity, the support vector machine(SVM) prediction model are used to predict the hot spot temperature of the transformer.

To address the problem that the number of UAVs is increasing dramatically and the collision risk is increasing dramatically, autonomous collision avoidance and collision avoidance path optimization for conflict risk among multiple UAVs by using genetic algorithms. Under the performance constraints of UAVs, genetic algorithm can realize autonomous collision avoidance of multiple UAVs by using adaptive crossover probability and adaptive mutation probability, and an adaptation function based on the shortest collision avoidance path length and the lowest collision risk. The results of the three-UAVs collision avoidance simulation show that the genetic algorithm can prevent the occurrence of chain collisions while solving the current flight conflict problem and guarantee the safety of UAV flight.

In this paper, we propose an efficient curb detection and trackingmethod (ECDT) based on high-resolution 3D-LiDAR.The ECDT follows manual-feature based technique routine, and the algorithm is consist by five steps. Firstly, the ECDT downsamples the raw point cloud and performs rough ground segmentation. Then we project the non-ground points into a density image under bird’s eye-view (BEV) and combine the improved beam model to realize the identification of road intersection and the classification of laser points. Thirdly, we down-sample the ground points and project them into BEV to obtain height difference image, and use gamma stretching upon the obtained image to extract candidate points. In addition, three filtering operation are applied to obtain accurate feature points. Finally, the algorithm perform curb fitting and tracking. In the experiment, we first compared the ECDT algorithm with the classic 3D curb algorithm, and verified the real-time performance of our method. Then we further test the ECDT under multiple complex scenarios by RSReference ground truth system. The results show that the proposed method has high accuracy and satisfactory robustness.

Lightweight design is a key way to realize the engineering application of deep learning algorithms on Unmanned Aerial Vehicle (UAV) platform. Aiming at the low detection accuracy of the current real-time object detection algorithms of the UAV platform, a lightweight model based on Vision Transformer (ViT) is designed in this paper. Firstly, a small Convolutional Neural Network (CNN) is used to extract primary features for reducing the number of parameters and computation amount of ViT network in this model, and using window modeling to replace part of the global modeling. Then, a feature-level mask self-supervised training method is applied to pre-train the ViT structure, which helps to accelerate the convergence and avoid a lot of labeling work. Finally, the result compared with other UAV lightweight object detection algorithms in the visdrone2018 dataset shows that this model has higher average accuracy on ensuring real-time speed, and verifies the effectiveness and reference value of the lightweight design method proposed in this paper.

The path planning algorithm is important for the safety and stability of self-piloting. In this paper, target bidirectionalRRT*(TB-RRT*) path planning algorithm based on target gravity and improved metric function is proposed to address the problems of low search efficiency, high randomness, slow convergence, and unsmooth path of the rapidly-exploring random tree star (RRT*) algorithm. Firstly, the algorithm introduces the target gravity and dynamically adjusts the sampling step to improve the search efficiency of the algorithm and reduce the randomness of the search tree growth; through the bidirectional tree growth strategy, the convergence speed of the algorithm is improved. Secondly, the smoothness of the planned paths is improved by considering both Euclidean distance and pinch angle effects on path planning with an improved metric function. Finally, the smoothing path with optimal path cost is obtained by the path selection method. Simulation results show that the improved TB-RRT* algorithm reduces the path length by 12.8%, the running time by 56.9%, and the number of sampled nodes by 43.8%.

Aiming at the low detection speed caused by the complexity of YOLOv4 target detection algorithm,a lightweight YOLOv4 algorithm is proposed.Firstly,the backbone of YOLOv4 is replaced by the lightweight network MobilenetV3 to reduce the number of network parameters;The 3×3 standard convolution in PANet is replaced by the depthwise separable convolution to reduce computational load;K-means is used to acquire the anchor box.And improved YOLOv4 is tested by using Udacity dataset.According to the result,we can find that under the premise of ensuring the detection accuracy, FPS is increased by 74% and the speed of the improved YOLOv4 is improved.

Deep learning-based panoptic segmentation technology has enabled in-depth understanding of images as well as scene descriptions and provides solutions for most computer vision tasks. The post-processing fusion of instance branches and semantic branches in the top-down approach often causes conflicts between branches. In order to solve this problem, we propose a single-stage panoptic segmentation network based on pixel-level instance perception with FCOS object detector. Specifically, the FCOS output extraction network is extended by using three branches, namely the object detection branch, the semantic branch, and the panoptic branch. The object detection branch uses the FCOS object detector to provide the instance center for the panoptic branch to generate a pixel-level instance perception mask, and the semantic branch provides semantical information for the panoptic branch. The panoptic segmentation output comes from the pixel-wise product of the semantic branch and the panoptic branch, avoiding the heuristic post-processing fusion in end-to-end learning. A good trade-off between accuracy and inference time was achieved on standard datasets Cityscapes, with PQ at 59.4%, mIoU at 76.2% and running time 94ms on Cityscapes validation dataset. Compared to other methods, our method achieves better panoptic performance within a reasonable runtime range. At the same time, the inference process is concise and can be optimized by using the deep learning inference engine to reduce the difficulty of deployment in actual application scenarios.

As awareness of network security being enhanced, more and more encryption technologies are used in cyberspace. The security detection of e-mail system in encrypted traffic becomes a challenge. This paper proposes an analysis system of email attack in encrypted traffic based on abnormal behavior, which identifies mail behavior and determines abnormal behaviors by extracting relevant feature information in the traffic. The special model and general model of the system were used to test the encrypted traffic of the existing network. The results show that the efficiency has been greatly improved compared with traditional detection methods.

In the context of Industry 4.0, AGV trolleys are increasingly being used. An AGV path planning method based on the logistic-tent chaotic grey wolf optimization algorithm is proposed for automatic guided vehicles (AGVs) with reduced population diversity, a tendency to fall into local optimality as well as poor solution accuracy and imbalance between global and local planning during path planning. Adding Logistic-tent chaotic mapping can initialize the initial population diversity of gray wolf individual locations. And changing the convergence factor control strategy of gray wolf optimization algorithm can improve global path optimality and speed up convergence. In this paper simulation results show that the logistic-tent chaotic grey wolf optimization algorithm has better solution accuracy and global search capability, and can quickly plan a better feasible path.

In recent years, the development of intelligent manufacturing systems around Digital Twin has failed to meet the digital needs of actual enterprises in key areas such as design, management, production, and service, and the integration of physical object data and twin data is generally low and failing to meet the requirements of rapid response to production regulation. In response to this problem, this paper proposes a digital twin-based product lifecycle information fusion method from the perspective of Product Lifecycle Management, eliminating the problem of information silos among systems, and completing multidimensional data fusion in three aspects: the creation of multidimensional models, construction of ternary data models, and integration of multi-system service information. Finally, project verification is performed in the actual production environment, and the integration of data is realized.

The dark channel prior is a statistical conclusion based on outdoor haze-free images and it is widely used in haze removal because of its effectiveness. However, it does not work well when dealing with images containing sky areas. The reason is values of R,G,B channels are very high in this region , i.e. the dark channel prior does not hold . Therefore, the transmission of this region is overestimated, resulting in serious halo artifacts and noise amplification in the recovered image. To solve this problem, an image dehazing algorithm based on the inverse channel compensation prior is proposed in this paper. According to observation and experimental verification, for haze-free images, when the pixels in the sky region of the image have higher values in a certain channel, their inverse channel values are lower even tend to be close to zero. Therefore, when finding the dark channel of the image, the inverse channel is considered for compensation so that the pixels in the failed sky region also satisfy the dark channel prior assumption. It is worth noting that the method based on the inverse channel compensation in this paper can obtain the accurate transmittance, and the sky region will not produce distortion and noise amplification even if the lower constraint on transmission is not set . In addition, a simple and effective color cast judgment and adaptive correction method is proposed based on the characteristics of atmospheric light , and adaptive color correction if necessary, which makes the method in this paper effective for the enhancement of sand dust images , extending the applicable scenarios of the algorithm.

Aiming at the time-varying, nonlinear and non-stationary problems of continuous non-invasive blood glucose detection data, a Gate Recurrent Unit (GRU) neural network based on particle swarm optimization (PSO) is proposed. Firstly, the electro-optic method was used to extract the Photoplethysmography (PPG) signal of the human body under red and infrared light, and the relevant characteristic parameters were extracted based on the physiological signal characteristics and the vascular elastic cavity model. Finally, a continuous non-invasive blood glucose detection model based on GRU was established, and then the parameters of GRU neural network were optimized by the PSO algorithm with strong optimization ability, which effectively improved the accuracy of the blood glucose detection model. Experiments have verified that GRU optimized by PSO has better accuracy and stability than the traditional neural network, and its accuracy reaches 89.3%.

With the continuous improvement of network accuracy, the depth and width of the network also increase, so the requirements for storage and operation equipment are also higher and higher, which limits the application scenarios of the network. To solve this problem, many network acceleration methods have been proposed, among which thresholdbased pruning algorithm and differentiation-based network architecture search algorithm are two commonly used methods. The optimal pruning threshold is found by searching through a differentiation-based network architecture search algorithm using an incremental approach. Experiments were carried out on CIFAR-10/100 using VGG16, ResNet-56 and ResNet-18 respectively. The experimental results show that the algorithm in this paper can help the network to maintain better classification accuracy after pruning with a higher compression rate.

Micro deformation monitoring technology can be used in surface subsidence, slope, building deformation detection and other emergency fields. A micro deformation monitoring method for multiple targets by exploiting a mmWave radar are proposed. The proposed method takes advantage of employing higher-order cumulant (HOC) to localize targets and extracting the phase by differentiate and cross multiply (DACM). The measured data in real scene have validated the proposed method. It is convenient to be applied in real engineering practice with a high accuracy and a low computing complexity.

The elevator balance coefficient is an important index to measure the safety operation of the traction elevator. The traditional measurement scheme of the elevator balance coefficient is time-consuming and laborious, and the improved measurement scheme is limited by the weakness of the traditional acquisition system, which leads to the popularization difficulty. In this paper, A multi-channel high-speed data acquisition system is designed by ZYNQ system core board, A/D acquisition module, SD card, and HD display screen. The system temporarily stores the collected data through FIFO, reads the data through DMA to PS terminal of ZYNQ, and stores and sends the data after processing by PS terminal. Finally, the QT platform is used to build a simple user interface to control the display of collected data. The experiment proves that the system realizes the high-speed acquisition, transmission, and display of multi-channel data, and has the characteristics of high acquisition accuracy and fast acquisition speed.

Aiming at the problems of inaccurate distribution network topology relationship, complex lines and lack of accurate topology information records, this paper proposes a topology identification method based on big data technology. By collecting the information of switching current, power and current harmonics at all levels in the station area, the electrical parameter identification metabase is established, and the clustering analysis of the metabase is completed based on the improved k-means algorithm, according to the results of cluster analysis, the topological connection relationship of equipment in the station area is obtained. Through verification and analysis, the proposed method effectively improves the accuracy of topology identification of distribution station area, realizes the efficient and accurate identification of topology of low-voltage distribution station area, and provides an important foundation for the intelligent application of distribution network.

In order to ensure that mobile robots perform various tasks correctly, the research on efficient and practical path planning algorithms is of great significance. For the Artificial Potential Field algorithm, it is easy to fall into local minimums, the target points are unreachable, and the A* algorithm search time is long when planning the moving path. The speed is slow, the planned path is not smooth enough and it is not the optimal path. This paper proposes a mobile robot combined path planning algorithm based on the APF algorithm and the A* algorithm. This algorithm can solve the local oscillation problem while reducing the problem. Obstacle avoidance time, improve the efficiency of the navigation algorithm. From the results of MATLAB Simulation, it can be seen that the algorithm can complete path planning accurately and efficiently. The mobile robot can bypass the local minimum point and successfully reach the target point, achieving the expected path planning effect.

Aiming at the problems of low efficiency and low accuracy of protocol analysis in the actual network environment with large real-time traffic and wide use of unknown protocols, we propose a novel fast format classification and parsing approach based on multiple progressive resolution (FCPMPR). Firstly, binary message is scanned by experience window, and the message is compressed by distribution entropy and word frequency. Secondly, for avoiding excessive information loss caused by dimensionality reduction, the Simhash algorithm is adjusted to map the indefinite length packets to fixed-length fingerprint. And the rapid classification of packets is realized by fingerprint distance. Then in the same type of message, a domain-supported dynamic time warping (DDTW) algorithm is proposed to match the message feature structure with different protocols and marking density centers. Finally, the over-divided feature structures are combined to realize structure and state words recognition. Experiments show that the proposed algorithm achieves both accuracy and time complexity effectively based on multi-layer progressive method.

Polyadenylation plays a decisive role in the life cycle of mRNA in eukaryotic cells. The primary sequences of cis-elements of plant poly(A) sites are much less conserved than those of animal sites. The secondary structure of RNA sequences tends to be more conserved than the primary sequence and is closely related to the function of mRNA, therefore it is necessary to investigate whether there is potential secondary structures around plant poly(A) sites. Here we proposed a pipeline to characterize secondary structures surrounding plant poly(A) sites. First, poly(A) sequences of Arabidopsis were converted into the secondary structure form represented by dot-bracket notation, which is then denoted by a position number matrix and converted into a triple structure. The triple structure is based on the structure of the stem region, on which each structural unit of the secondary structure is described. Each structural unit has a corresponding mathematical model that is used to determine various structures of the sequence. Finally, the size and position of secondary structures surrounding poly(A) sites are obtained. Using this pipeline, we provided a detailed view of the secondary structure of plant poly(A) regions, which help understand the functional importance of plant poly(A) signals and associated secondary structures.

In recent years, the research of data visualization has attracted more and more attention, especially the visualization of traffic flow data. This paper proposes a grid clustering method based on k-means. First, the geographic data is divided into grid matrix, and then K-means is used to perform cluster analysis on the data. The traffic flow data is then displayed. In this paper, the inner feeling of public transport flow analysis is very important for public transport operation decision making, and provides a very important basis for the operation safety supervision and risk assessment decision of urban public transport.

Facial landmark detection is challenging with occlusion, pose or inadequate training samples. We proposed a two-branch facial landmark detection network (Facial Detection with Face Structure and Transformers: FDFST) considering face structure constraints. Existing regression-based facial landmark detection models have not fully considered the general facial structure for landmark detection, that usually lead a unstable prediction. In contrast to facial landmarks, facial structure is more likely to be accurately estimated in the real scenario. Therefore, we try to provide a facial structure guidance for the facial landmark detection by a facial structure estimation sub-network. In this way, two targets are predefined to supervise our model, one is the facial structure described by five landmarks, the other is facial landmark denoted by 96 points. To address the lack of occlusion samples, we proposed a novel data augmentation to boost the training process on the public data sets WFLW. Experiments have revealed that our FDFST network on the WFLW dataset achieved significant improvement.

How to improve the detectability for tiny faces is a major challenge in face detection. To solve this issue, we raise a single-stage high performance network for tiny face detection. In feature fusion, multi-scale contextual feature selection and spatial pyramid pooling modules are adopted to extract contextual information with a fixed proportion to mitigate the information loss in the highest-level feature map of feature pyramid. In addition, during training, an improved anchor compensation strategy is used to match enough anchors for tiny faces to promote the learning of tiny faces. A test is conducted in the standard data set FDDB. Our method reaches 98.5% TPR (true positive rate) when the amount of false positive faces are the same as 1000. Meanwhile, a test is conducted in the standard data set WiderFace, and the terms of AP reach 97.0%, 96.2% and 91.8% on the easy, middle and hard subset. The algorithm in the paper can improve the performance of the face detector effectively.

Fire is one of the most frequent and common disasters endangering public security and threatening social development. There are various causes of fires, but fires that cause major personnel and property losses are usually caused by untimely detection, late alarm, and other reasons. In the initial stage of fire, the traditional smoke-concentrated sensor is often used to detect the temperature of the fire. If immediate and effective measures are not taken in the early stage of fire, it is easy to cause large personnel and property losses. Therefore, aiming at the problem of smoke detection in the early stage of fire, an embedded smoke recognition system based on object detection is proposed in this paper. The system adopts data enhancement to expand the data set, transplants the YOLOv5 object detection algorithm to the embedded system to recognize smoke, and sends the alarm information to the terminal. It has good recognition effect and meets the needs of smoke detection in the early stage of fire.

With the advancement of unmanned driving technology, unmanned vehicles' application environment has gradually moved from the structured road surface to the unstructured and complex field environment. The traditional path planning approaches cannot match the path planning needs of unmanned vehicles in the field environment due to the complicated and changing terrain. In this paper, a fast random tree search algorithm is used to find a viable path in an environmental map quickly. By introducing the target probability bias value and the path post-processing method based on the maximum turning radius constraint of the unmanned vehicle into the basic RRT algorithm, and then simulating and comparing the simulation analysis on the 3D raster map of the constructed field environment. The applicability of the RRT algorithm based on the target probability in the field environment is verified, and the algorithm has higher search efficiency, smoother paths planned by the algorithm, and more goal-oriented path generation.

With the rapid development of wireless positioning technology, high precision has gradually become one of the important goals pursued in the positioning field. The existing distance-based UWB ranging technology has low positioning accuracy and even lower accuracy when it is blocked by obstacles. This paper proposes a spatial ranging method based on the combination of PSO-SA algorithm and KNN, aiming at the accuracy of positioning solution and the problem of signal interference in the positioning process. First, use the basic commands of Matlab to grab the measured distance in the data file, and remove the vacancy and the same distance data, then remove the unreasonable distance values in the data according to the Laida criterion and establish a polynomial fitting equation according to the characteristics of the UWB hardware system Second, establish a UWB spatial positioning model and solve the model through the PSO-SA algorithm; then, for the obstacle occlusion problem in the positioning process, the KNN model is used combined with the ranging information of each UBW anchor point to assist in judging the obstacle occlusion situation and establish a new model. Finally, the model obtained by the positioning method is verified by experiments. Experiments show that the algorithm has a 3D accuracy of 98.79% in an indoor environment with obstacles, a 98.59% positioning accuracy in a 2D XY plane, a 1D accuracy of 97.75% in the x direction, and a 1D accuracy in the y direction is 97.92%, and the 1-dimensional accuracy in the z-direction is 90.62%. The algorithm can improve the positioning accuracy performance of UWB technology, effectively eliminate measurement errors, and achieve accurate positioning when there are obstacles.

The detection of small and dim targets in infrared image sequences is a key technology in infrared search and tracking systems. The optimization method based on low-rank and sparse decomposition is the mainstream research method in this field. Most of the existing optimization methods are based on a single frame, only exploiting spatial information but ignoring the time dimension. The multi-frame detection method makes full use of temporal and spatial information to achieve higher accuracy. However, the existing multi-frame detection method is very slow, because they rely on slow decomposition methods. To solve the problem, we propose the infrared sequence tensor model for multi-frame detection. First, we select the tensor average rank to describe the low-rank property of the infrared sequence. It can be solved quickly. Second, we introduce a new approach to extract prior information as a prior weight tensor, which can highlight the target and suppress noise and strong edges on the background. Third, we formulate the optimization equation based on the tensor average rank and prior weight, which can be solved efficiently with tensor robust principal component analysis(TRPCA). Experiments show that our proposed method has high detection accuracy. The speed of our proposed method is much faster than that of the current multi-frame optimization methods and is comparable to that of the singleframe optimization methods.

An adaptive beamforming algorithm based on undesired disturbance cancellation and steering vector constraint is proposed to solve the problem of beamforming performance degradation caused by formation distortion and array model mismatch during passive sonar operation. This algorithm reconstructs interference-plus-noise covariance matrix (INCM), which is closer to the real value through the uncertainty set constraint of the steering vector, discusses the composition and estimation method of the undesired disturbance of Capon power spectrum, and gives the basic implementation flow of the covariance matrix reconstruction algorithm. The robustness of the algorithm is verified by simulation results.

In view of the characteristics of the current non-invasive blood viscosity calculation that is complex and easy to be interfered, resulting in low accuracy of calculation results,to further improve the accuracy of blood viscosity prediction,in this paper, the machine learning algorithms of extreme learning machine and random forest are used to accurately measure blood viscosity.First of all, we need to extract the PPG signals of different people and preprocess the PPG signal waveforms to obtain high-quality PPG waveforms. At the same time, the pressure pulse wave signals are collected and preprocessed.Secondly, extract the feature points of the two waveforms and calculate the pulse wave feature parameters according to the extracted feature points.The blood viscosity value is preliminarily estimated according to the characteristic parameters, and the value and other human parameters are used as the input parameters of the prediction model of blood viscosity,through extensive training of parameters, the best prediction model is selected, thereby improving the prediction accuracy of blood viscosity.The experimental results show that the predicted value of blood viscosity obtained by the random forest algorithm is better than that obtained by the extreme learning machine algorithm, and the accuracy reaches 88.4%.

Super-resolution(SR) is a classic issue in computer vision. Existing SR methods mainly rely on artificially downsampling to prepare the pair-wised training data set without considering the blur kernel inconsistency and often result in an unstable performance in the real scenario. Especially for facial images, unacceptable reconstruction seriously affects its downstream applications. In order to improve the quality in real facial image SR, we proposed an end-to-end deep neural network to mimic a realistic process of image degradation and reconstruct a high-resolution (HR) facial image from a low-resolution (LR) one. At the first stage, a generative adversarial network (GAN) is employed to sample more realistic LR images by jointly considering facial landmark and composite blur kernels. At the second stage, pairwised training samples are fed into a decoder to fit the mapping from LR to HR. Experiments were performed on real LR test set from Widerface and LS3D-W dataset. Results demonstrate that our model outperforms existing competing methods. Furthermore, the ablation studies valid the effectiveness of the proposed components.

In the mining and analysis of time series data, the similarity measurement process of two time series is essential. The dynamic time warping algorithm DTW (Dynamic Time Warping) is a common algorithm for measuring the similarity of time series. Although DTW has obtained the global optimal solution, it may not be able to achieve local reasonable matching, and the measurement effect is limited. The shape dynamic time warping algorithm shapeDTW (Shape Dynamic Time Warping) enhances the measurement effect by considering point-wise local structure information. In order to continue to improve the measurement effect of the shapeDTW algorithm, this paper analyzes the definition and common methods of the shape descriptor of its algorithm. On the basis of extending the subsequences at each time point, the method of representing subsequences using shape descriptors is directly changed. The DTW value of the subsequence is calculated as the distance value of each time point in the cumulative cost matrix during the alignment of the entire original time series. This method is called subDTW (Subsequence Dynamic Time Warping). The subDTW algorithm enhances the measurement effect of the shapeDTW algorithm to a certain extent. Through the experimental research and analysis of the approximate window size of the segmentation aggregation on the training set, the subDTW algorithm based on the cumulative cost matrix pruning can keep the classification accuracy unchanged. A significant reduction in time overhead is achieved.

Object counting is an essential topic in computer vision due to its practical application of surveillance analysis. Most existing works focus on single-view or multi-view object counting and have achieved good performance. However, stacked objects appear frequently in industrial site and these counting is viewed as a difficulty. In this paper, a new counting method of approximates cubic stacked objects (e.g. stacked material) is proposed. The approximate cubic stacked objects count by fusing two vanishing points and single-view object counting. Two vanishing points are used to provide partial stereo information and single-view object counting is used to provide quantity information. We perform experiments of stacked object images captured in actual industrial sites. Experimental results show that our method is effective in approximate cubic stacked object counting.

This paper intends to improve the dissemination of Chinese Chaozhou woodcarving by text mining analysis with users’ comments on TikTok. By LDA topic modeling analysis, we found the main themes users discuss are about: 1. Chaozhou woodcarving skills and aesthetics. 2. Chaozhou local cultural recognition and self-confidence. 3. Inheritance and continuation of Chaozhou woodcarving. Based on the analysis of user discussions on World Intangible Cultural Heritage, with Chaozhou woodcarving as an example, we suggested that in order to improve the dissemination, we need to: highlight the regional characteristics, arouse the hometown feelings and cultural resonance of young people in the Chaozhou area, add more practical elements to the work, and let it develop from art works to practical items.

This paper collects three singing methods of Hua'er 《 White peony Ling 》 and analyzes four acoustic parameters: duration, energy, fundamental frequency and long-term average spectrum.In terms of duration, the longest pronunciationduration of cry singing is mainly due to the long pause between its sections.From the perspective of energy distribution, the energy value of crying singing is greater than the other two, and the energy dispersion is also greater. Fromtheperspective of fundamental frequency distribution, the peak value of fundamental frequency from vast mountains singingmethod to crying singing method to cheerful singing method gradually changes to the low-frequency area, and fromthevalue of standard deviation, it is gradually changing. The dispersion of fundamental frequency of vast mountains singingmethod is large.From the long-term average spectrum, the change forms of LTAS curves of the three singing methodsare very similar, and the spectral envelope characteristics are similar. The average LTAS curve position of cheerful singing method is the lowest, and it is considered that there is a singer formant in Hua'er singing.

Aiming at the problems of short text length, less information contained in text and poor classification performance caused by easy interference of deep learning model, combined with the advantages of CNN and BiLSTM, a text classification method based on fusion semantics, attention mechanism and BiLSTM-CNN model is proposed. Firstly, the Skip-gram model in word2vec is used to map the words in the data and convert them into vectors with fixed dimensions as the input of the model. The local features of the text vector are extracted by CNN and the global features related to the text context are extracted by BiLSTM. The features extracted by the two complementary models are fused, It solves the problem that the single convolution neural network model ignores the semantic and grammatical information of words in the context, and effectively avoids the gradient disappearance or gradient dispersion of the traditional cyclic neural network. This paper makes an experimental analysis on R8 and Ohsumed datasets. The results show that the proposed feature fusion model effectively improves the accuracy of text classification.

In order to overcome the low accuracy of traditional endpoint detection algorithms in low SNR speech endpoint detection. In this paper, EMD is used to decompose manic speech adaptively, and multi-window spectrum subtraction is used to reduce the noise of each decomposed signal to improve the SNR of speech signal. Based on the MFCC principle, the improved MFCC spectrum entropy was proposed, and then the product of logarithmic energy and MFCC spectrum entropy was taken as a new parameter, and median filtering was performed on it. Two dynamic thresholds were designed for this new feature parameter, which could better track and locate the start and end of the actual speech. Matlab simulation results show that compared with other common algorithms, the detection accuracy of the improved algorithm is improved by 2.2% under the -5dB Gaussian white noise, and by 2% under the -5dB Volvo noise. Not only from the perspective of innovation, but also from the perspective of performance, it has good accuracy and certain robustness under the low SNR environment.

Now the mortality of cancer is high, which seriously threatens the safety of human life. It is very difficult to accurately segment small objects such as cancer lesions or cells. In order to improve the accuracy of small object segmentation, this paper proposes to use RPN module to frame and select small objects, then use FSRCNN to enlarge the low resolution picture, and finally use a simple U-shaped segmentation model to segment small objects. The model has been evaluated on the cell data set and achieved good results.

UAV-borne interferometric SAR has unique advantages and has attracted the attention of key research institutions in recent years. Baseline is a crucial parameter in InSAR processing, which is directly related to the accuracy of InSAR measurements. Compared with space-borne InSAR, UAV platform is difficult to keep the flight at the same track, and the flight attitude is unstable, which brings great difficulties to the baseline estimation of UAV-borne InSAR. In this paper, an improved baseline estimation method is proposed. This method derives the relationship between the interference phase period, the slant range, the distance between two ground points, and the phase difference through the interference geometric relationship. The distance between the two points is then substituted into the known parameters for baseline estimation using the least squares method. Compared with the original method, the proposed method can fully exploit the potential of the interferogram data by applying the least square method to solve the formula between flight parameters and baseline parameter to achieve baseline estimation. The simulated results and the flight test results of the original method and the proposed method are Compared. The absolute error, RMSE and standard deviation of both methods are calculated. It shows that the accuracy and robustness of the proposed method are significantly improved.

The odometer (OD) had been widely employed for the land vehicle navigation with an inertial navigation System (INS) during global Navigation Satellite System (GNSS) outages, but the navigation error of such system still accumulates with the measurements from the odometer. In this paper, we present a Map-Matching (MM)-based approach to enhance the positioning performance of the micro-electro-mechanical systems (MEMS) INS/odometer integrated system during GNSS outages, without additional equipment. More specifically, the MEMS INS and odometer are integrated based on an extended Kalman filtering (EKF), and then the location sampling points are entered into MM module based on trajectory segmentation criterion to predict the position and azimuth on the road network, and finally the error correction is completed through EKF. The main contribution of this study is to point out the essential cause of the position and velocity error drift of INS/OD system through observability analysis, and to suggest a novel hybrid strategy to integrate the MEMS INS, odometer based on both Kalman filtering and Map-Matching, which inhibits divergence through position and azimuth observation provided by map. The proposed navigation strategy is verified via road tests with artificial GNSS outages ,showing that the root mean square of azimuth angle error, horizontal velocity error and horizontal position error of the proposed navigation strategy decrease by 93.2%, 92.6% and 95.5% when the 200s GNSS interruption time is set.

Transmission line sensing equipment can realize real-time measurement and monitoring of transmission line body operation status and environment, which is a necessary prerequisite for disaster warning and improving transmission line operation and maintenance. With the in-depth development of line maintenance work and the need for the construction of strong smart grid, the application scenario of transmission line field sensing technology is becoming more and more extensive, and the sensing means and equipment are increasing day by day. This paper summarizes and analyzes the principle and status of the current field application of perception technology, discusses some common problems of perception technology, and looks forward to the future development direction of this field.

Auxiliary power supply system is an important part of medium and low speed maglev train. It is mainly composed of auxiliary inverter, charger, suspension power supply, several battery packs and several loads. Taking Qingyuan medium and low speed maglev train as an example, the actual topology of its auxiliary power supply system is analyzed, the models of auxiliary inverter, charger and suspension power supply are established in the simulation system, and the internal working principle and working process of the system are studied in detail. The auxiliary inverter adopts three loop control and SVPWM control mode of current and voltage double closed loop, which improves the stability of the auxiliary inverter during load switching. Combined with the charging characteristic curve of battery pack, the constant current and constant voltage charging control strategy is studied, and the control logic of the simulation system is designed according to the actual operation of the auxiliary power supply system. The system level simulation on MATLAB®/Simulink® platform shows that each component can work stably under rated parameters, and meet the design requirements and the working requirements of auxiliary power supply system. Finally, its rationality and stability are analyzed. The robustness of the auxiliary power supply system model is verified by simulation. This research provides a certain theoretical basis for the research of auxiliary power supply system of medium and low-speed maglev train.

In order to improve the accuracy of the common cause failure (CCF) analysis results of Aviation Ground Lighting(AGL) power supply system, an improved integrated part method (UPM) is proposed. Taking a medium-sized airport light station black-out accident (SBO) as the top event, a system fault tree is constructed. Calculate and analyze the Fussell-Vesely importance of minimal cut sets in fault trees. Identify CCF groups and build a system fault tree that includes common cause failures. UPM was applied to determine the beta factor value for each CCF group. The comparative analysis considers the occurrence probability of important system events before and after CCF. The results show that the application of the UPM-CCF analysis method more truly reflects the reliability level of the lights power supply system.

The inter-turn short-circuit fault in rotor windings has a destructive impact on the wound rotor synchronous motor. It is necessary to identify the fault timely and accurately in the early stage of the fault in order to avoid serious losses. Most of the existing fault diagnosis methods are sensor detection for a single fault feature, which is difficult to ensure high diagnosis accuracy for a long time. This paper analyzes the multi-source fault characteristics of inter-turn short-circuit fault in rotor windings, and can accurately identify the fault type by relying on a variety of signals. Firstly, this paper analyzed the time domain and frequency domain characteristics of motor voltage and current performance and rotor under unbalanced magnetic tension. Then, considering the actual running state of the motor, the accurate model of the motor is established using the finite element simulation tool, and the multi-physical field coupling simulation is carried out for the fault. The multi-source fault information characteristics are analyzed from the perspectives of electromagnetism, heat and force. Compared with various information of the normal motor, the motor rotor fault can be identified quickly and accurately.

The electric energy router is produced in order to meet the requirements of various power supply forms, multi-directional energy flow and active power flow control of the new power system. It is based on numerous power electronic devices, which brings harmonic problems. This paper proposes a dead zone compensation strategy based on electric energy routers. The method can compensate the power quality degradation issue caused by the dead time of the switching device of the power router and improve the stability of the power router connected to the power grid. The main task of this paper is to consider the negative effects of dead time on power router harmonics, grid-side current, etc. Firstly, the basic structure of each module of the power router is introduced, and the control strategies of different modules are introduced. Then, the dead time mechanism is introduced in detail and the influence of dead time on power quality is explained by Fourier decomposition of error voltage waveform. Finally, the principle of dead zone compensation is introduced and applied to the power router of this paper and the power indicators after compensation are briefly analyzed. In the subsequent simulation verification, this paper analyzes effects of different dead time on harmonics and power quality indicators, and the simulation results verify that when the dead zone time is increased, the power indicators of the power router are getting worse and worse. Finally, the effectiveness of the dead zone compensation is verified by comparing the power indicators added to the dead zone compensation strategy.

In this paper, we investigate the constellation design for the multiple-input single-output (MISO) dimmable visible light communication, where a total average- and a peak-intensity constraint are imposed on the channel input. An optimized spatial constellation is obtained by an equiprobable equally-spaced signal constellation (with a carefully-chosen minimum distance) concatenated with the minimum-energy signaling mapper. Error performance and rate loss of the proposed spatial constellation at high SNRs are evaluated as well. Numerical results show the performance advantage of our proposed constellation over the benchmarks.

The key to a low-carbon integrated energy system lies in the combination of carbon capture power plants and integrated energy. During the peak load, one of the characteristics of carbon collection plant - high energy consumption and the anti-peaking characteristics of wind power, there is a low-carbon coordination problem. In view of this drawback, first of all, the source side utilizes the smart operation mode of carbon-collecting power supply plants. Secondly, considering the system uncertainty, an optimization model aiming at optimizing the energy consumption of system using integrated energy is constructed. Finally, simulation verification is carried out in the improved system of the carbon capture power plant to achieve the goal of improving the absorption capacity and operating return of the integrated system while achieving the goal of low carbon emissions.

This paper proposes a sea surface radar echo simulation method based on wave spectrum, which generates data from the generation mechanism by calculating the radar echo intensity at various points on the simulated sea surface. The method can generate echo intensity data matching the set sea state and radar parameters without the actual measured sea clutter data. The simulation results show that the simulated clutter waves fit well with the measured sea clutter and can be used as an effective sea clutter simulation method.

In the optical wireless communication system, a reliable transmission channel can be achieved in the power restricted scene by using a single-photon avalanche diode (SPAD) receiver. However, the performance of SPAD is limited by dead time. This results in count loss and inter-slot interference (ISI), which degrades the communication performance. In this paper, a 4-channel time-gated scheme was proposed. In the scheme, time-gated SPAD receivers successively turned on by external gating signals to reduce the effective dead time of the receiving system and improve the signal-to-noise ratio. In addition, the 4 time-coordinated receivers scheme has better background light tolerance and better performance in intensive signal power condition than the 4 free-running SPADs receivers scheme.

With the wide application of the hybird gas insulated switchgear (HGIS) in the new power system, the very fast transient overvoltage (VFTO) and transient enclosure overvoltage (TEV) generated by the operation of disconnector in HGIS have brought serious harm to the insulation status of equipment, which is one of the main threats to the safe operation of the power system at present. This paper takes Delingha 800kV HGIS substation in China as the research model and uses the electromagnetic transient simulation software-ATP to calculate and analyze the numerical results of VFTO and TEV under different operating modes. The simulation results show that, compared with the traditional model, the overvoltage amplitude of the HGIS enclosure is as high as 0.1-0.2 p.u. in the model considering the electromagnetic transient process of the HGIS enclosure. The amplitude and frequency of VFTO and TEV generated by multiple power supply - multiple outlets operation mode are higher compared with single power supply - single outlet operation mode, and the amplitude of VFTO and TEV generated by the disconnector closer to the power side is higher.

When high-penetration new energy is connected to the grid, it has a huge impact on the frequency stability of the system. In this paper, an adaptive control strategy for hybrid energy storage based on state-of-charge recovery is designed. Firstly , the operation timing and range of each kind of energy storage is determined according to the different frequency regulation dead band of the two kinds of energy storage and their respective SOC ; Secondly , energy storage frequency adjustment is divided into frequency modulation period and SOC recovery period according to whether the frequency deviation is greater than the frequency modulation dead band . During the frequency modulation period , the action depth of virtual droop and virtual positive/negative inertia is determined by SOC and frequency deviation ; In the period of SOC recovery , a self recovery method of SOC is proposed based on the grid limit coefficient determined by frequency deviation and the energy storage demand coefficient determined by SOC . Finally, by establishing a frequency modulation model to conduct simulation experiments on different types of interference, it is verified that the proposed strategy can effectively maintain the stability of the system frequency.

As lightning strikes greatly threat the safety of transmission line, the locating of lightning strikes is of great significance to fault recovery and safe line operation. Distributed optical fiber sensing technology has the characteristics of strong anti-electromagnetic interference ability and wide monitoring range, and has broad application prospects in the field of online monitoring of transmission lines. This paper proposes a lightning strike locating method based on ΦOTDR. Through simulation, the impact of lightning arc force and shock wave on the cable and its internal stress changes are analyzed. Through the lightning strike experiment in the laboratory, the ΦOTDR data during lightning strikes are analyzed, and the results show that using ΦOTDR, the vibration situation near the lightning strike can be monitored, and the signal-to-noise ratio of the vibration signal can reach 6.3dB after EMD decomposition-wavelet decompositionwavelet denoising processing, which provides a new method for lightning strike monitoring of transmission lines.

Our country's southeastern coastal provinces are relatively developed economically and consume a lot of electricity, but these provinces generally have the problem of not being self-sufficient in energy. With the development of technology, long-distance DC transmission technology is usually required to supply part of the electricity. In a regional grid with DC input and internal generator sets, various loads, and various new architectural forms, its transient stability has also changed accordingly. In order to improve the reliability of power supply under the new situation, it is necessary to study the characteristics of such local power grid frequency, voltage, and power angle under fault conditions, and to provide a control method that can improve system stability.This paper will focus on the transient characteristics of the local power grid under the above conditions, and combine the virtual synchronous generator(VSG) control strategy to improve the transient stability of the system.In the latter part of this paper, the transient stability of the system is analyzed combined with a simulation example, and the effectiveness of the control strategy is verified.

Millimeter-wave radar is widely used in autonomous driving. A MIMO antenna array and corresponding motion compensation method are designed to solve the Doppler-angle coupling problem of time-division multiplexing multipleinput multiple-output (TDM-MIMO) frequency-modulated continuous wave (FMCW) radar. First, overlapping-elements are introduced into the traditional MIMO antenna array, and the phase difference between the overlapping-elements is used to eliminate the motion-induced phase errors. Then, the phase error caused by the position error of the antenna element is corrected by an iterative method. Finally, the FFT algorithm is used to accurately estimate the moving target's distance, velocity, and angle. Simulation experiments show that the method can effectively solve the Doppler-angle coupling problem, improve the angle estimation accuracy of moving targets and avoid the multi-target matching problem.

The stability issue of power system has raised more attention these years since the typical characteristics of traditional power grid have changed significantly, due to the increasing proportion of renewable energy and power electronic interfaces. For wind power generation, it is faced with the problem of oscillation in a wide range of frequency, and the oscillation reflects the characteristics of frequency coupling. This paper firstly carries out theoretical analysis and simulation verification on its characteristics and generation mechanism, and concludes that the oscillation replication characteristic of PLL is one of the influencing factors.Therefore, proposes a new control strategy for the wind farm with VSC-HVDC transmission system, which weakens oscillation characteristics of the phase-locked loop, reduces the risk of instability caused by the coupling of oscillation frequencies. Finally, the control strategy is simulated and verified by MATLAB/Simulink software.

Electrical impedance tomography (EIT) is a non-invasive medical imaging technique, and its reconstruction process has an ill-posed inverse problem. This paper firstly bases on a B-spline curve electrical impedance imaging shape reconstruction model, and secondly obtains the training data of the network by using the EIT forward problem. Finally, image reconstruction is performed using the Auto-Encoder method. The data obtained by simulation and experiment verifies the effectiveness of the method, and the method has good adaptability to the measurement data, the imaging quality of the boundary shape is high, and the imaging efficiency is also improved to a certain extent.

Aiming at the stability requirements of high-precision current sources in such areas as inertial navigation, this paper designs a high-stability current source. We analyze and compare the Howland current source scheme in advanced foreign instruments and the current source scheme used in our design and explore the key parameters that affect the stability of the current source output. The compensation circuit is designed and analyzed, and the hardware diagrams are proposed. The results show that the 10mA and 1mA current output stability of the current source can reach 1ppm (24 hours),which meets the design requirements.

Due to the increasing penetration of new energy generation based on power electronic devices in the power grid, the characteristics of the power system are changing. As a new and continuously developing concept proposed in recent years, the grid-forming(GFM) inverter can adapt to the grid with high penetration of renewables, especially can provide support for the power grid in the weak grid. In this paper, the synchronous power control(SPC) strategy is adopted , and the behavior of the GFM inverter is analyzed during the fault. To address the overcurrent problem under weak grid, a current-limiting strategy is proposed considering the effect of the power outer loop, which limits the current within the allowable range while maintaining the grid structure in the weak grid, and improves the dynamic response. Finally, a simulation model is built in PLECS to verify and analyze the proposed control strategy.

The prediction of electric vehicle charging load can provide an important basis for power grid planning and design. Aiming at the problem that the existing methods are insufficient in the prediction accuracy of electric vehicle charging load, a Kernel Extreme Learning Machine (PSO-KELM) electric vehicle charging load prediction model based on Particle Swarm Optimization (PSO) algorithm optimization is proposed. The model adopts the Monte Carlo method to simulate the electric vehicle charging load for a week by considering the electric vehicle ownership, charging mode, initial charging time and driven distance, and uses PSO-KELM to predict the charging load. Experiments show that the prediction model can comprehensively consider the influencing factors of charging load, effectively improve the accuracy of electric vehicle charging load prediction, and provide a reference for the optimal operation and planning of the power grid.

Under the condition of power market, how to coordinate the medium and long-term trading plan with the power generation dispatching is a new problem. Considering the medium and long-term trading electricity, this paper proposes a daily generation planning method based on standard load curve. The power plants are divided into priority power plants with balanced provincial regulation, market-oriented power plants without regulation capacity and marketoriented power plants with regulation capacity; The first type of power plant curve adopts the predicted value, the second type of power plant determines the power generation curve according to the unit of different standards in the flood and dry seasons, and obtains the standard load curve from the load curve of the whole network, and the third type of power plant adopts the standard load curve to determine the power generation curve. The results show that this method can effectively connect the medium and long-term trading electricity with the daily power generation plan.

Due to the emergence of a large number of endogenous hidden dangers, defects and loopholes of network security, the network security threat faced by power monitoring system is becoming more and more serious, which puts forward new requirements for monitoring, operation and maintenance personnel. Based on the construction goal of the security operation coordination platform in the power industry at the present stage, this paper proposes to enhance the network security operation system of full view defense and in-depth defense from the aspects of management, technology, capability and coordination. Build a security operation and maintenance coordination platform for the power industry, integrate technologies such as situation awareness, investigation and evidence collection, scheduling and linkage disposal, and solve the problem of insufficient power emergency response and safe operation capacity. It has improved the basic safety monitoring ability and safety protection ability of the power industry. Make the safety protection level of the industry move forward from static perception and active defense to dynamic perception and intelligent defense.

Subsynchronous oscillation (SSO) can severely threaten the safety and stability of power systems. To study the mechanism and reduce the harm of SSO, the measurement of SSO current is of great significance. However, SSO current is comprised of three components with extremely close frequencies, i.e., a power-frequency component, a subsynchronous component, and a supersynchronous component, which makes it difficult to separately measure the components’ accurate values. To address this issue, this paper provides a novel time-domain method for measuring SSO current. The proposed method contains two key parts: 1) a multi-stage transfer circuit (MTC) comprised of two cascaded stages, where the SSO current serves as the input signal; and 2) a smart SSO current measurement algorithm (SMA) to give the accurate values of the above components, depending on the signals output by the stages of the MTC. The SMA takes advantage of the transfer characteristics of the MTC and the symmetry of the sinusoid signal to calculate the component parameters, which successfully separates the components with the close frequencies. Also, the proposed method can be easily implemented. Moreover, simulation tests have been carried out, and the test results verify the effectiveness and accuracy of the proposed method.

Due to the large distributed capacitance of EHV transmission line, its relay protection scheme is obviously different from the traditional method, and the existing method cannot realize the whole line speed. This paper analyzes the basic method of zero sequence current differential protection, discusses the influence of capacitor current on the current protection of EHV transmission line, and puts forward the zero sequence current innovation protection method. When the zero sequence current innovation exceeds the threshold, the starting criterion locates the fault line. This method is not affected by fault location and fault type, and improves the reliability of protection algorithm. The feasibility of the protection algorithm is verified by PSCAD simulation.

With the increasing proportion of distributed new energy in the grid, more and more battery storage systems are also connected to the grid to improve the negative impact of new energy generation on the grid. Energy storage systems are mostly connected to the power grid through grid-connected inverters, and a large number of grid-connected inverters will have a great impact on the stability of the network. Therefore, this paper adopts impedance analysis method and stability analysis based on Nyquist criterion to study the interaction between grid-connected inverter and power grid of energy storage system. Based on the current mode control algorithm of energy storage system, the impedance model of energy storage inverter is established, and its operation stability under different conditions is analyzed. In order to improve the grid-connected stability of the energy storage inverter under the condition of weak power grid, an active damping suppression strategy based on input current feedback was adopted to improve the harmonic stability of the gridconnected energy storage inverter under the condition of weak power grid. Finally, the accuracy of the analytical conclusions and the effectiveness of the proposed method are verified by simulation results.

The optical storage DC microgrid is taken as the research object. In view of the general situation of unstable output of photovoltaic cells, the energy storage system is introduced to absorb power in case of excess photovoltaic output power and release power in case of insufficient photovoltaic output power, so as to maintain the power balance of microgrid. The stability of bus voltage in DC microgrid is an important index of power balance in the system. Therefore, maintaining the stability of DC bus voltage is an important way to maintain the power balance of microgrid and system stability. Aiming at the problem of bus voltage stability, a method based on single neuron PID adaptive control is proposed. Taking bus voltage as the control variable, the influence of single neuron PID controller on Bidirectional DC-DC converter is studied. The results show that the single neuron PID adaptive control method has a good effect on controlling the voltage stability of DC bus, enhances the anti disturbance ability of the system and improves the operation stability of optical storage microgrid.

In view of the current provincial and municipal power supply companies in the process of allocating investment funds, there is a problem that the distribution of investment funds is roughly based on analysis of historical data, which is easy to form investment imbalances and cause weak links in power grid construction to be unable to be reasonably strengthened.This paper proposes an evaluation and analysis method of power grid investment benefit efficiency based on DEMATEL-improved grey relational method, and constructs a power grid investment benefit efficiency index system from the perspectives of construction necessity, economy, safety and power supply reliability. The evaluation laboratory method (decision making trial and evaluation laboratory, DEMATEL) accurately identifies and selects the key influencing factors, and uses the improved grey relational analysis method based on information entropy to quantify the correlation degree of the influencing factors to find the weak points of power grid investment and construction link. In the case analysis, the ten key factors affecting the investment efficiency of the power grid were screened out by the method, and the comprehensive scores and construction weaknesses of different power supply areas were obtained. The situation proves the effectiveness and rationality of the method, which can guide the investment direction scientifically and improve the investment efficiency.

A lightning forecast method based on potential forecast is proposed to solve the problem that only radiosonde data and potential forecast can not accurately predict lightning prone areas. By collecting transmission channel, the history of urban distribution network lightning observation data, combined with the static mesoscale numerical model forecast products with technology and real-time altitude detection data analysis, through comparative analysis of grid real-time lightning observation after correction, finally get of the potential forecast more accurate than traditional lightning is 0 ~ 72 h potential forecast results. The research results realize the active defense of lightning damage in power grid, guarantee the safety of power grid and reduce the economic loss caused by power failure to users.

In recent years, wireless power transmission has been greatly developed, and it has been used in commercial applications. The research on wireless charging of implantable medical devices is also deepening. This paper mainlyanalyzes the magnetic field distribution of the asymmetric coils in implantable medical treatment and the coils withdifferent offsets in actual medical conditions. The transmission power based on Parity Time Symmetric circuit isintroduced. Finally,simulate the transmission power of implantable medical equipment, and it is demonstrated that thetransmission power can be stable under different offset conditions.The results can be applied in the field of wirelesscharging in implantable medicine.

Aiming at the problem of excessive configuration capacity in multi-energy microgrid (MG) planning, this paper first analyzes the system optimization of MG, establishes the flow chart of the optimal sizing of the entire system, then proposes the optimal sizing constraints considering the load loss probability, and establishes a configuration planning model, selects the appropriate configuration method for solving. By using particle swarm optimization (PSO) algorithm to calculate and analyze, the optimal sizing result can be obtained. Furthermore, simulation is conducted combined with case studies to compare the configuration selection of peak load, and verify its economy through simulation.

The examination and management of mental stress has been recognized as a vital examination scheme. However, present methods for stress-assessment rely mainly on questionnaire surveys, which can provide the changing trend of a characteristic value under stress state, but fail to determine the specific stress level. To tackle this problem, this study developed a non-invasive and objective stress measurement system signals collected by biosensors. A wearable sensor was used to collect PPG signal in real-time, and the short-term heart rate variability was measured, and SDNN and LF/HF were extracted as characteristic values in the time and frequency-domains to detect mental stress. In order to examine the performance of the proposed system, physiological signals of 20 volunteers under stress and normal states were collected. The stress state was identified according to the stress-state table of each characteristic value. The proposed system resulted in a stress-prediction accuracy of 95% in distinguishing between low- and high-stress levels. The results indicated that SDNN and LF/HF are effective indicators for the detection of mental stress, and can effectively improve the accuracy of stress-recognition.

In today's society, the power supply network often needs to carry out uninterrupted power transmission, and the demand for live work is getting higher and higher. Manual live work has its difficulties and limitations, especially the increasing complexity of modern transmission lines and equipment, which increases the safety risk of manual work. Through the live work robot of the distribution network, the difficulties and limitations of manual live work can be overcome, thereby replacing manual work. The article first expounds the necessity and market prospect of live robots in distribution network, and then introduces the domestic and foreign research status of key technologies for live robots in distribution network, and then briefly describes the structure of hydraulic manipulators, motion planning algorithms, hand-eye calibration methods, and anti-collision. Algorithms and five key technologies of live working robots for distribution networks in the visual servo control system. Then, according to the actual operation process of live robots from Yi Jiahe Technology Co., Ltd., the main research difficulties and improvement strategies currently faced by live working robots in distribution networks are summarized. The future development trend of robot technology for live work in distribution network is prospected.

With the wide application of distributed generation technology, in order to meet the requirements of grid adaptability test of distributed system. A voltages control strategy based on Fuzzy-PI and quasi-proportional resonant (QPR) is proposed. Based on the traditional voltage outer loop PI control, fuzzy control is added to realize the real-time update when the system is disturbed and improve the stability of the system. It can improve the dynamic response and anti-interference ability of the system quickly when it is disturbed by external factors. MATLAB/Simulink was used for experimental simulation. According to the experimental results, the control strategy designed has a good tracking effect, which reduces the total harmonic distortion rate of the system and improves the steady-state performance of the system.

The power market is gradually transitioning to the mode of "power generation side quotation and user side quotation" and the stage of joint clearing of electric energy and auxiliary services. Therefore, it is necessary to study the power market in which interruptible loads participate in joint clearing under the bilateral quotation mode. In this paper, the risk cost caused by interruptible load's participation in the spot market before the day is analyzed and studied, and a joint clearing model of electricity energy and reserve auxiliary service market in the spot market before the day is established. The members of the market on the load side are introduced into the market in the form of quotation, and an example analysis of a provincial power grid proves that the mechanism of load's participation in the joint market proposed in this paper is superior.

With the rapid growth of the communication system infrastructure of The State Grid and the rapid advancement of the digital economy in the digital era of the power grid, it has become an urgent need for the rapid development of China's energy industry to find the key technologies to realize the digitalization of the security management and network intelligence of the State grid. Digital twin technology is also leading the development of the power grid digital era. Through digital identification, intelligent sensing, data resource pooling, digital twin modeling, and other key technologies, it effectively leads the safe and stable operation of the entire power grid in China. Firstly, this paper focuses on the historical development background, connotation and characteristics of digital twin technology. Combined with digital twin technology, this paper analyzes the development status of Smart power grid and digital power grid in China, and then introduces the key technologies based on digital twin technology in the field of power grid development, as well as the typical applications of these technologies in power grid enterprises. Then it points out the technical difficulties in the construction of digital twin power grid and the challenges in the field of power grid. Finally, it provides important theoretical support for the long-term construction and development of digital twin power grid and points out the new direction of power grid development.

In this paper, a sensor access system based on streaming computing mode is studied and implemented to meet the requirements of fast access and real-time processing of large-scale sensing nodes, which provides fast access and management capabilities for sensors. Aiming at the service carrying capacity of nodes after access, the time characteristics of network traffic are studied, and the optimal path algorithm is designed based on this feature to ensure the stability after access. The test results show that the application sensitive to delay and jitter can obtain more bandwidth resources to ensure the quality of service and improve the performance of multi-service bearing based on the priority fast access mode and the optimal routing algorithm.

In sensorless control system, High-frequency (HF) voltage injection methods are widely used for permanent magnet synchronous motor (PMSM). However, subject to the response of rotor estimate position, either the speed loop or current loop bandwidth can not be set too large. But the current loop bandwidth is a key factor to accuracy control. With low bandwidth, it can only handle DC and low-frequency disturbance. Because the dead-time of voltage source inverters (VSIs) which cause the 6th harmonic in the d-q coordinate system, a revised repetitive controller (RRC) is proposed to reappear specific frequency harmonics. This paper described a new injection method, which combines the HF voltage and RRC output. In limited bandwidth, the 6th harmonic in current is sharply decreased. Finally, simulation results have proved the feasibility of the proposed scheme.

In order to solve the problem of inaccurate force load prediction results, a power load prediction method is proposed.By acquiring the historical power load data of different monitoring points, maximize the minimum normalization of the historical power load data, construct the power load input matrix based on the normalized power load data, construct the convolutional neural network to extract the spatial features, construct the gated cycle unit to introduce the attention mechanism, input the extracted spatial features to extract the time features, construct the periodic features of the power load data, and integrate all features for prediction.

The damage of stator armature winding inter-turn short circuit fault to motor determines the necessity of reliable and effective diagnosis in the early stage of fault. Most of the existing fault diagnosis methods are based on single fault feature detection by sensors, which is difficult to ensure long-term diagnostic accuracy. Based on the multi-source information fusion theory, this paper conducts a multi-physics coupling simulation of the stator fault by constructing the finite element model of the motor under different working conditions and analyzes the change of the multi-physics characteristics of the motor stator under the inter-turn short-circuit fault. Using electromagnetic, thermal, mechanical, and other characteristic parameters to diagnose stator winding faults can effectively and accurately identify fault types, and greatly improve the reliability of synchronous electronic stator winding fault diagnosis.

In the stage of rapid design of the aircraft, the flight load has an essential influence. According to the flight parameters required for flight load analysis, a nonlinear flight dynamics model is established in this paper. The longitudinal extreme maneuver action is designed and optimized based on the national military standard. The pitch angle rate controller is designed based on the L1 adaptive control method to ensure the stability of the aircraft. After the flight dynamics model is combined with the extreme maneuvers, the control law is embedded for simulation. The results show that the UAV can complete the extreme maneuver simulation under L1 adaptive control, and the overload can meet the expected target. The correlation between the control system and the extreme maneuver simulation is small when the control parameters are within the appropriate range.

On the basis of the dynamic tracking control of quadrotor UAV, combined with the dynamic characteristics analysis and automatic detection of UAV flight targets, the stability and accuracy of dynamic tracking control of UAV are improved, and the design method of dynamic tracking control system of UAV based on Arm-Linux embedded platform is proposed. Using machine learning method and acceleration feedback method, the target object model of quadrotor UAV dynamic tracking control is established, the quantitative characteristic parameters of quadrotor UAV dynamic tracking control are analyzed, the hardware processing platform of UAV dynamic tracking control system is established under Arm-Linux embedded platform by adopting dynamic feedback adjustment and ambiguity detection, and the nested saturation control method is adopted in the target visual distribution of tracking control objects. Combined with video image feature detection method, the visual feature detection model of quadrotor UAV dynamic tracking control is established. Based on normal visual guidance image and anti-interference design of dynamic tracking control, the dynamic tracking control of quadrotor UAV is realized and the system design is completed. Simulation results show that this method has good stability and strong visual tracking ability for UAV dynamic tracking control, which improves the dynamic stability and balance of UAV.

To achieve carbon peaking and carbon neutrality goals under the new development philosophy, environmentally friendly autonomous driving has become one of the focuses of the automobile industry. It is essential to optimize the fuel consumption in the modeling studies conducted on the trajectories of autonomous vehicles. Moving bottlenecks caused by low-speed vehicles may influence traffic efficiency and generate extra fuel consumption. The emergence of connected self-driving technology makes communications among vehicles and better trajectory control possible. This paper investigates the effects of cooperative adaptive cruise control on fuel consumption under a moving bottleneck. The moving bottleneck is caused by the overtaking of trucks on a two-lane expressway. Our analysis results indicate that cooperative adaptive cruise control can efficiently reduce fuel consumption under the moving bottleneck compared with Gipps’ carfollowing model. Also, the optimization efficiency of cooperative adaptive cruise control increases as the control distance decreases and the initial speed increases.

This paper proposes a tightly coupled GNSS-VIO system based on Visual, IMU, and GNSS measurements. For complex areas, the pseudorange proves fast and efficient, but can only achieve meter-level positioning precision. The carrier phase can achieve centimeter-level positioning precision but is required to solve the ambiguity of the whole cycle, which is difficult and time-consuming. To combine the advantages of the two measurements, we replace the pseudorange factor with the carrier phase smoothing pseudorange factor in the factor graph. In addition, to obtain accurate positioning results, we need to calibrate the extrinsic translation parameter between GNSS and IMU. Therefore, a real-time calibration method based on smoothing pseudorange is designed. Experimental results on public dataset show that our system not only eliminates the drift of VIO system but also has better positioning accuracy compared with the existing loosely coupled and tightly coupled GNSS-VIO systems.

Actuator saturation has an important influence on the stability of the quadrotor UAV (Unmanned Aerial Vehicle) control system. The characteristic of the control structure of the quadrotor UAV is that the inner loop control commanddirectlyaffects the actuator. Aiming at the problem of actuator saturation of quadrotor UAV attitude control, an attitudecontroller based on Conditioned Super-Twisting Algorithm (CSTA) is designed. The two sign functions in CSTAarereplaced by the hyperbolic tangent function, and the input of the hyperbolic tangent function is adjusted byagainparameter, thereby suppressing the chattering of controller. The experimental results show that the designed controllercan track the attitude command better than PID when the actuator is saturated. Compared with the CSTAalgorithm, thedesigned controller output chattering is significantly reduced.

When the air compressor of the bus compresses the air, water will be generated and condensed in the air reservoir. Water will rust pneumatic components until they fail. Therefore, drainage is required. The water level sensor and drain valve of general electric drainage system are easy to fail. The automatic drainage system designed in this study cancels the water level sensor and adopts regular drainage. Add electric condenser. At the same time, the drainage valve is improved. It not only meets the needs of drainage, but also enhances the reliability and service life of the system.

Aluminum conductor steel reinforced (ACSR) will produce broken strand defects after a period of operation, which seriously affects the safe and stable operation of transmission lines. In this paper, a two-dimensional LGJ-240/30 type ACSR broken strand defective heat transfer model is proposed to study the corresponding relationship between the number of broken strands of ACSR and the temperature rise of the defective backside and the load current. The model is simulated using COMSOL Multiphysics to obtain a fitting formula for the number of broken strands under natural convection conditions. The test platform is built under natural conditions to verify the feasibility and accuracy of the formula. The results show that, using the formula proposed in the article, the number of broken strands of ACSR can be identified more accurately under the condition that the load current and backside temperature are known, and improve the efficiency of infrared inspection.

In today's increasingly stringent emission requirements, the requirements for aftertreatment are also increasing. Controlling the motor in the post-processing system plays a key role. In this paper, the method of controlling the motor will be analyzed. When analyzing the traditional PID controller, it is found insufficient, so a new single neuron PID controller is used, which can better control the DC brushless motor than the former; the latter structure has better adaptability. The advantages and disadvantages of the two controllers are compared in the Simulink simulation of MATLAB, and it is concluded that the latter has a faster response speed and stronger adaptability.

In this paper, the power generation costs of different DG and public energy storage in DC microgrid (MG) interconnected system are analyzed. On this basis, a decentralized nonlinear dynamic droop gain control strategy based on equal micro-increase rate criterion is proposed, which automatically adjusts the output of DG to make the microincrease rate of each DG and two MGs in the same MG equal to public energy storage, so as to achieve economic optimal operation. Finally, the decentralized economic optimal operation control strategy is verified in the simulation.

The brushless DC motor is used as a vector control system of the robot. Using the magnetic field oriented control scheme of Hall sensor, the hardware circuit design and FOC software algorithm design are given. The simulation model of motor drive system is built based on MATLAB / Simulink software platform, and the parameters of motor electromagnetic torque, speed and voltage are observed and analyzed. The experimental results show that the designed control system is reasonable and feasible, and has high cost performance.

This paper presents a design method of simulation result record base, which includes simulation result recording module, simulation result management module, simulation result analysis module and simulation result playback module. The simulation result record base records and manages the simulation results of system comprehensive evaluation, provides a comprehensive management platform for simulation data, and also provides data support for equipment system performance evaluation, analysis, report, simulation playback, equipment model verification and calibration in the later stage.

Satellite application simulation support software is used for the simulation of satellite applications such as satellite communication, positioning and navigation and reconnaissance. This paper proposes a design method of satellite application simulation software. The software consists of satellite scenario editing and three-dimensional visual display unit, satellite orbit simulation unit, satellite orbit simulation unit, satellite link simulation unit, satellite networking communication simulation unit, satellite reconnaissance and surveillance simulation unit and satellite navigation and positioning simulation unit.

Brushless DC motor (BLDCM) is characterised by time-varying, non-linear, strong coupling and many variables, and have been extensively studied in the area of motor control. The traditional PID control method as a regulator is not ideal to achieve the desired speed system control effect. Therefore a fuzzy PID controller is designed to implement. The working principle and mathematical model of brushless DC motor (BLDCM) is briefly analyzed, and a simulation model of the brushless DC motor control system is built in the Simulink module of MATLAB software. Speed closed-loop speed regulation is realized by fuzzy PID control, while the fuzzy PID control is used to achieve the speed closed-loop speed regulation, and a double closed-loop speed control system model of the BLDCM is built. Finally, through the analysis of the simulation results, the fuzzy PID control BLDCM drive system has faster dynamic response and stronger anti-interference ability compared with the traditional PID control.

Storage technology, Peer-to-Peer network technology, Peer-to-Peer storage system and other related technologies are discussed. A peer-to-peer storage system is built by using peer-to-peer network technology combined with the hard disk capacity on the network nodes. It uses file segmentation technology to increase the utilization rate of storage space, and parity check code and Reed Solomon code files Fault-tolerant technology to greatly improve the availability of files. File availability is the key index of the system, which is mainly affected by the number of file partitions, backup methods, fault tolerance and other parameters. In this Peer-to-Peer storage architecture, storage device hosts can achieve mutual backup effects, thereby improving the reliability of the system.

Aiming at the problems of cumbersome wiring, high power consumption, low precision, and low signal-to-noise ratio in the practical application of the electrochemical hydrogen sulfide sensor in coal mines, a wireless sulfid e sensor based on Wave-Mesh networking wireless networking module and C8051F021 microcontroller was desi gned. The hydrogen sensor completes signal amplification, analog-to-digital conversion and software filtering thr ough the internal integrated circuit of the single-chip meter, and uses the idle sleep mechanism inside the singl e-chip microcomputer to complete the low-power processing of the whole machine, combined with the Wave-M esh networking wireless networking module. Wireless transmission. Through the basic error and power consump tion experiments, the results show that the power consumption of the sensor can be controlled within 49mAh, the sensor accuracy meets the requirements of industry standards, and the continuous and reliable monitoring of hydrogen sulfide in coal mines can be realized.

This paper presents a simulation method and software for the interaction between environment and equipment mobility, including the simulation model of the interaction between natural geographical environment and equipment mobility performance, the simulation model of the interaction between man-made battlefield environment and equipment mobility performance. The simulation environment elements can simulate equipment trafficability, collision detection and passage, which can effectively support equipment simulation.

A multi-path feature fusion network model and data augmentation algorithm are proposed based on the limitations of the remaining useful life prediction method of a single network model and the problem of small sample size. Firstly, the sample size of data set is expanded through data augmentation, so as to improve the accuracy of prediction. Secondly, two different paths are selected to extract features: the first path inputs the data into the cascade structure of 1- Dimensional Convolutional Neural Network and Gated Recurrent Unit, and extracts spatial and temporal features respectively; the second path inputs the data directly into the Long Short Term Memory to obtain the complete temporal features. Finally, the output features of the two paths are fused, and the result input into the full connection layer for RUL prediction. The proposition is evaluated on the publicly available health monitoring dataset C-MAPSS of aircraft turbofan engines provided by NASA. The results show that compared with the single network model, the proposed method can effectively improve the remaining useful life prediction accuracy of such equipment, and has practical application value.

During the working process of the rice-wheat combine harvester, the irregular accumulation of grains on the scraper of the scraper-type grain elevator will affect the final measurement accuracy. In this paper, a scraper type combine harvester grain flow monitoring device is designed. In this paper, the input capture function of STM32 single-chip microcomputer is used to obtain the high-level signal output by the array photoelectric opposite-beam sensor, and the grain accumulation information on the elevator scraper is obtained; the discrete element simulation software is used to analyze the grain accumulation model on the scraper, and the calculation formula of grain flow is established; the principle of BP neural network is used to improve the monitoring accuracy of grain flow. The indoor bench test and field dynamic performance test were carried out. The results show that the opposite-beam grain flow monitoring device designed in this paper operates stably, and the maximum relative error of the field harvesting test flow monitoring is 4.12%, which meets the actual needs of combine harvester grain flow monitoring, which provides a technical basis for field grain flow monitoring.

With the development of computer technology, all kinds of instrument operation have higher and higher requirements for data processing, and all kinds of intelligent measuring instruments have come into being. As one of the typical representatives of intelligent measuring instruments, digital voltmeter has only a history of more than 50 years since its inception, but there have been electronic tube type, transistor type, various scale integrated circuit type, intelligent type and so on. With the progress of technology, the production cost of digital voltmeter is getting lower and lower, the stability is improving, and the range is also expanding. Various advantages brought by the development of science and technology make digital voltmeter favored by all walks of life. The measurement parameters of digital voltmeter can only measure a single parameter, but now it can measure a variety of parameters, such as DC voltage, AC voltage, impedance, power and so on. This paper plans to design a three channel AC / DC digital voltmeter with simple structure and convenient operation. It can realize the automatic measurement of three channel AC and DC basic parameters. After many considerations, it is determined to design this digital voltmeter based on single chip microcomputer.

Most of the servo systems use a three-loop nested control structure of position, speed and current, in which the closed loop of position and speed mainly depends on the position signal of the photoelectric encoder and the speed feedback based on it. This paper considers how to use the encoder with limited precision to improve the control performance of the system as much as possible under the premise of system stability, and also reduce the cost of the control system. Firstly, the encoder position feedback error is analyzed and compensated, and then two different speed estimation methods are used to obtain the speed feedback based on the compensated position signal. Finally, the performance of the proposed method and traditional estimation methods are compared on an actual commercial servo drive platform.

The planar magnetic induction angle sensor made by PCB process is more resistant to impact and vibration than the traditional photoelectric encoder, but the absolute position coding method nonius coding is not easy to realize in the high-precision sensing unit. There are a great many pole pairs of high-precision sensing units, resulting in a small error correction range of cursor coding and easy angle jump, so there are high requirements for the installation of sensors. To solve the above problems, a planar magnetic induction angle sensor based on m-sequence coding is studied in this paper. The coarse code channel of the sensor adopts differential m-sequence coding, which effectively improves the absolute coding error correction range of the sensor. The actual measurement of the prototype shows that the sensor can realize reliable absolute position measurement under loose installation conditions, and the measurement accuracy is ± 10 "in the range of 0 ~ 360°.This coding method not only improves the error correction range of the sensor, but also provides a new idea for the coding method of planar magnetic induction angle sensor.

The incident report is a common approach used by the international ATC industry for safety management. The report analysis may lead to inconsistent results or missing risk items due to subjective judgements of human. Based on identification of hazards, this paper proposed an automatic method for classifying incident reports by using natural language processing from a safety risk management perspective. The Runway Incursion(RI) reported events were retrieved. Firstly, the hazards related to RI were identified, and report text was calibrated. Then, the similar text was generated to enhance the text data. After that, the words of text were vectored. Finally, the incident reports text classification model was trained. The results show that this method can automatically classify RI reported events according to the categories of hazards.

In order to improve the yaw motion stability of the rear compartment of modern trackless train connected by six articulated devices, a yaw rate feedforward-feedback control strategy is proposed. First, the feedforward desired yaw rate is calculated based on the 3 degree-of-freedom linear reference model, which includes the first two compartments of the vehicle. Second, the problem of calculating compensate yaw rate is solved by sliding mode algorithm, and the target yaw rate is obtained by adding the feedforward desired yaw rate to the compensated yaw rate. At last, the PI control algorithm is used to calculate the yaw moment which is distributed to the wheel by the torque distribution algorithm. U-turn and S-turn cases are designed to verify the performance of the control strategy. Simulation results show that the proposed feedforwardfeedback controller has less yaw rate tracking error compared with the uncontrolled one, and the path-following error is less than 0.12m. The yaw motion stability of the vehicle is guaranteed.

With the development of the Internet, People's life and work depend on the network for longer and longer. All kinds of network data are forwarded through the network, and the network information flow is increasing. These network data occupy more storage space and consume more energy to forward these network data. With the continuous development of intelligent technology, intelligent requires real-time network response. Therefore, the busy network conflicts with the development of intelligent technology. It is urgent to improve the efficiency of data forwarding without affecting the network.

Since the action mechanism and properties of muscle relaxants are affected by many factors, in order to achieve the safety and rationality of anesthesia, it is necessary to correctly assess the degree of muscle relaxation of patients. Therefore, an easy-to-operate acceleration muscle relaxation monitor was designed. . The acceleration muscle relaxation monitor is based on 51 single-chip microcomputer as the main control unit, and the main body is divided into switching power supply, pressure sensor, A/D conversion module, display screen and printer. The acceleration muscle relaxation monitor is divided into four stimulation modes, which can complete a variety of stimulation combinations. The acceleration muscle relaxation monitor can complete the task of monitoring the patient's muscle relaxation index in real time, and reduces the occurrence of intraoperative and postoperative muscle relaxation residuals. And the equipment is simple in design, convenient in operation, and easy to popularize.

As we all know, the power system shows a trend of diversification and innovation, the circuit breaker control circuit, as an important link in the power system to ensure the normal circuit breaker, directly affects the quality of power system operation. Once the control circuit is disconnected, technicians are required to fully analyze the cause of the fault and take corresponding measures to ensure that the circuit breaker can be opened and closed normally. However, it is difficult for field operators to find the problem due to the lack of professional knowledge. In this paper, based on the analysis of the control circuit break defects and elimination in Chuzhou area in recent years, and combined with the control circuit break principle research, an intelligent diagnostic instrument for circuit breaker control circuit disconnection is designed, which can effectively help field operators to find the cause of the fault, and provide maintenance measures for common fault causes.

In order to solve the problem of water accumulation or corrosion, pneumatic brake is widely used in the braking system of commercial vehicles. Use the dryer to remove the moisture and oil in the air of the brake system, so as to ensure the clean and dry air in the air pipe. Exhaust valve integrated design improvement reasonable structure, durable and reliable, can effectively improve the service life of the product.

With the wide application of online conference in plenty of companies and institutions, the maintainers need to monitor the real-time video stream of the main venue and branch venue to ensure the success of conference. In order to improve the maintenance efficiency of State Grid, this paper designs an intelligent conference video monitoring and management system that integrates various image processing technologies and artificial intelligence algorithms with the cooperation of software and hardware. The system adopts the Qt development framework to complete the system interaction logic and display interface design. PaddleOCR tool is used to identify baffle text in branch venue picture, and YOLOv5 algorithm is adopted to detect people in the picture. Dlib library is used for participant attendance of main venue. The test and verification show that the system realizes the user login and management function, decoration guide and exceptional malfunction detection function of branch venue, participant attendance function of main venue. The maintenance requirements of State Grid can be met properly.

The converter of the distributed power supply is not similar to the damping component and rotational inertia of the conventional synchronous generator, which is not conducive to the stable operation of the power system.Virtual synchronous generator control technology simulates the same or similar external characteristics as the synchronous generator, which provides strong support for the stable operation of the power system. Considering the operation characteristics of the energy storage system, it proposes the virtual synchronous generator control method considering the energy storage operation characteristics. builds A simple AC microgrid simulation system is build, and the proposed control strategy verified.

The main research work of this paper is: based on a 6-DOF industrial robot model, the finite element software ABAQUS is used to analyze the modal characteristics of the model. Modal analysis gives the multi-order natural frequency and the corresponding vibration diagram. Based on the simulation results, the key natural frequency are found, which provide an effective reference for the design and optimization of industrial robot.