This PDF file contains the front matter associated with SPIE Proceedings Volume 12173, including the Title Page, Copyright information, and Table of Contents.

Two-dimensional (2D) materials have attract increasing interest in the past few years because of their good photoelectric, thermal and mechanical properties. In this paper, the photodetector of the two-dimensional material is taken as the main research object, at the same time, four kinds of photodetectors based on two-dimensional materials and their research status are introduced in detail. The methods of increasing the light absorption rate and improving the preparation of twodimensional material are put forward to improve the performance of photodetector. We fabricated a photodetector demo by MoS₂, and it shows a relatively high photoresponsivity and fast response time. Finally, the development prospect of two-dimensional material photodetector is prospected.

High power and high beam quality 2μm mid-infrared laser based on holmium doped lithium yttrium fluoride (Ho: YLF). Combine with gain switching technology, it can obtain stable short pulse width 2μm pulsed laser. Based on the Ho³⁺ energy level theory, through the analysis of the rate equation of the Ho: YLF quasi two-level system and the relevant parameters of the laser, the theoretical models of relaxation oscillation mode and gain switching mode under continuous pumping are established. In the experiment, 1940nm thulium (Tm³⁺) doped fiber laser is the pump source. Designed the structure of a semi confocal resonator with the flat-concave stable cavity. Under the 40W continuous optical pump, the optical efficiency is 30% and the slope efficiency is 76%. Under the same power, pulse pump the laser relaxation oscillation curve, and by compressing the pump pulse width the first pulse of relaxation oscillation is extracted. The relaxation oscillation and pulse image can well correspond to the theoretical model. The pulse parameters are 2050nm laser output with a pulse width of 2μs and single pulse energy of 500μJ.

In order to meet the data requirements in the development of space-based optical observation platform and simulate the debris motion tailing phenomenon in space-based optical observation images, a simulation algorithm of debris motion tailing in space-based optical observation images is proposed. Firstly, the imaging principle of CCD sensor is analyzed, and the simulation process of space-based optical observation image is designed. Secondly, the causes of space debris motion tailing are described. Finally, the imaging angular velocity of space debris is deduced, and the simulation method of space debris motion tailing is given. Simulation results show that the algorithm can truly simulate the motion tailing of space debris, which is helpful to the development of space-based optical observation platform.

We present a novel structure of planar waveguide for Yb:YAG laser amplifiers allowing large absorption length. The structure parameters of the planar waveguide were optimized using an 3D amplification model. The performance of the optimized planar waveguide laser amplifier was simulated, and a comparison between the cases with and without considering pump saturation was carried out. The simulated results show that a high pump absorption efficiency and optical-to-optical efficiency can be expected in addition to a good absorption uniformity. The seeder power is scaled from 200W to 7000W, the corresponding pump absorption and optical-to-optical efficiencies are 98.6% and 68%, respectively. The thermal stress of the designed planar waveguide is analyzed theoretically, the results show that the pump power of 10kW is available without the fracture risk.

In order to obtain high output power of 2 μm laser, the Tm³⁺ -doped fiber is pumped by a high power fiber-coupled diode laser, two fiber brag grating are used as the laser resonant cavity. The Tm³⁺-doped fiber laser operates at a maximum output power of 11.25 W and the slope efficiency is 28% compared to the pump current for 50 A. And the threshold is about 5.6 W. By increasing the pump power, the signal light power can be amplified, at the same time, an effective cooling device is designed to protect the doped fiber. When only one pump module is operating, the laser output power increased 19.3% after taking the device. On this basis when welding three pump modules, the laser output power is 36 W compared to the total current for 138 A.

In recent years, deep learning for mobile object detection has received increasing attention, and at the same time,handdesigned neural networks for mobile object detection have made remarkable progress, there is still a wide scope for exploring how to ensure accuracy and speed while pursuing the design of networks with fewer parameters and less computation, i.e., small and fast. Inspired by biological science, we know that humans only need to gaze at a glance, their visual system can have a rich perception of the complex environment and achieve a rough understanding of target detection, and the realization of this result relies on the corresponding prior knowledge, we propose a strategy of feature extractor crossover to adjust the results of the detection obtained by NAS with hyperparameters to achieve improve the detection performance on mobile devices.

In this paper, we propose an efficient method that combines the MobileNetV3 high-speed and the improved LSTM with temporal memory, and this network can effectively capture the key parts of the scene. Experiments show that we achieve an improvement in accuracy while ensuring the most important speedup in mobile. We reduction in the number of parameters from 4.9million to 3.96million, while reaching 90+ FPS in the Samsung Note10+ phone.

We proposed a graphene-assisted ultra-thin silicon electro-optic modulator. The optical permittivity, mode distribution, and propagation mode effective index were analyzed. The TM mode shows a higher power distribution efficiency in the cladding in which the graphene layers are placed. The modulator can work in electro refractive and electro absorption manners. While the real and imaginary part of the index are used for modulation, it has a low voltage-length product VL of 1.9 Vmm and a mode power attenuation of 3.27 dB/μm, respectively.

Analytical solution of the controllable self-accelerating and self-decelerating Airy-Ince-Gaussian wave packets in strongly nonlocal nonlinear media is investigated theoretically by solving the (3+1)D Schrödinger equation in cylindric coordinates. The ratio of the input power and the critical power, the initial velocity and the ellipticity effect the solutions.Thecontrollable self-accelerating and self-decelerating Airy-Ince-Gaussian wave packets are constructed by the Ince-Gaussian(IG) beams and the Airy pulses with initial velocity in temporal domain and in spatial domainrespectively.

Electrodes have an important influence on the terahertz (THz) radiation efficiency of photoconductive antenna, especially photoconductive antennas combined with nanostructures. The nanostructured photoconductive antennas with rectangular electrodes, traditional electrodes and bow-tie electrodes were simulated by the finite difference time domain method. The simulation results show that the bow-tie electrode is better than the rectangular electrode and the traditional electrode. The research results provide a promising scheme for selecting the appropriate electrode structure of the nanostructured photoconductive antenna.

When adaptive optics is applied to target imaging, laser atmospheric transmission, etc., variable extension target wavefront detection is a technical challenge. And its detection accuracy directly affects the correction effect of adaptive optics.In order to explore the wavefront detection accuracy problem of variable extension targets. In this paper, a simulation model for wavefront detection of variable extension targets with correlated Hartmann was developed. The model was based on the commonly used optical system parameters, and it was also established by using the theories of Fresnel diffraction, Newton's imaging equation, the working principle of Shack-Hartmann wavefront detector, and wavefront recovery.We analyzed the effects of target distance and attitude variations on wavefront detection accuracy. The relationship curves between RMS and PV of wavefront recovered residuals and target distance and different attitudes were obtained in the simulation. And we further carried out the analysis for the intrinsic physical reasons of forming this relationship. The results show that target distance and attitude changes affected the extension of targets within the Hartmann subaperture. When the wavefront was recovered using the correlation algorithm, we obtained that the smaller the extension, the higher the wavefront detection accuracy within the relevant Hartmann detection accuracy. Therefore, it could be summarized that the smaller the extension of the target, the more similar it was to the point target, then the detection error introduced by the extension became smaller.

In order to analyze and interpret the reflection image of the wedge splitter on the reflection surface of the laser terminal during the adjustment process, the mechanism of the double image and multiple image of the wedge splitter during the adjustment process was analyzed. Through theoretical analysis, a single wedge-shaped spectroscope fitting interpretation method is obtained, and the dominant relationship between the measured values of theodolite and the wedge Angle is quantified. The calculation method of the target value of the two wedge-shaped spectroscopes is obtained, which makes the interpretation accuracy of the reflection image of the two wedge-shaped spectroscopes increased to 0.5 ", ensures the system index and improves the production efficiency.

In the process of Few-Shot detection, the traditional data enhancement method of Few-Shot expansion is used in most cases. However, in the process of sample enhancement, some object images will be generated which are quite different from the real object. This brings about the error detection caused by expanding the sample. Therefore, this paper proposes a target correction technology based on Siamese Network to offset this kind of situation. Based on the Siamese Network framework, this method adds a second decision module to the output end of the target detection algorithm to adjust the detection result and improve the detection effect under the condition of appropriately increasing the detection time cost. In particular, the error target and Few-Shot target in the detection results have a large deviation problem, and further improve the accuracy of the target detection algorithm under Few-Shot. The test results on UCAS-AOD and other public data sets show that the algorithm improves the accuracy of target detection algorithm for small samples.

Polarization-dependent nonlinear absorption in the Co²⁺:MgAl₂O₄ crystal was investigated by the pump-probe technique. The nonlinear transmission curves of the [100]-cut Co²⁺:MgAl₂O₄ crystal for the linear polarized probe light were measured experimentally, when the crystal was excited by a linear polarized pump light. The measurement results showed that the transition into the near-infrared (NIR) absorption band was not only polarized along the C₃ local symmetry axis, but also along the direction perpendicular to the C₃ axis. And the absorption intensity of the latter transition was much weaker than the former. The structural model of Co²⁺ centers in the MgAl₂O₄ crystal, in which the Co²⁺ ion and surrounding O^2- ions formed a tetrahedron deformed along the C₃ local symmetry axis, was proposed. This structural model allowed us to explain the anisotropy of nonlinear absorption in the Co²⁺:MgAl₂O₄ crystal.

Room-temperature cascaded mid-infrared (MIR) pulsed laser output in Er:YAG crystal is reported. The 1469nm characteristic wavelength of the cascade emission is observed. The characteristic wavelength of the excited-state absorption (ESA) is determined as 1676nm. An Er:YAG crystal with concentration of 10at.% is adopted to compare the laser output energy in the cascaded and non-cascaded configuration. The maximum single pulse energy of MIR laser increases from 1.04mJ (non-cascade) to 1.51mJ (cascade), corresponding to a ratio of 45.2%. The “cyclic cascade” is considered to be the mechanism responsible for the laser performance optimization. The experimental results confirm the existance of cascade at room temperature in Er:YAG crystal. Moreover, cascade is helpful to improve the single pulse energy of MIR laser.

In order to solve the problem that the artillery pointing accuracy test can only be carried out in large outdoor field for many years, the autonomous north-seeking method and the north-seeking method with known north direction are studied first. Then, two-theodolites north diversion model and three-theodolites north diversion model are established. After that, the indoor north reference seeking method is given by combining north diversion models. Finally, the experimental verification shows that the method can easily and quickly establish the indoor north reference to meet the requirements of artillery pointing accuracy inspection.

Luminance and lifetime are key indicators for quantum dot light-emitting diodes(QLEDs) applications. Improving the hole function layer is a more effective method in many strategies for developing high-stability QLEDs. In our research, it was confirmed that QLEDs which use copper thiocyanate (CuSCN) has better performance than using traditional materials of poly(3,4-ethylene dioxythiophene): poly(styrene sulfonate) (PEDOT: PSS). The energy level of the device after using CuSCN is more matched, the transmittance of the red light band is higher, and making the hole injection capacity stronger. Luminance and service life have been increased by 53% and 68.5%, respectively. This work has far-reaching significance for the preparation of low-cost,and high-stability QLED devices that contribute to commercialization.

In order to explore a laser mirror system, the mirror driven by a uniform rotating motor can make the reflected laser scan linearly in the scanning plane, and overcome the disadvantage of traditional galvanometer scanning application. A two-dimensional model based on elliptical cylindrical mirror is established. The process of elliptical rotation causing the deflection of the reflected beam is analyzed. The general formulas of elliptical rotation angle and laser scanning point displacement in the polar coordinate system which rotate synchronously with the ellipse are summarized. According to the formula, MATLAB is used to analyze the nonlinear error between the elliptical rotation angle and the displacement of the laser scanning point in the elliptical cylindrical mirror system. A program using iterative algorithm is designed by MATLAB. Under the constraints of the relevant mirror size parameters, laser incident height range and scanning range given by the user, the program can find the mirror system that minimizes the nonlinear error of laser scanning and output its relevant parameters. At the same time, the relevant parameters of the linear equation between the rotation angle of the mirror and the displacement of the laser scanning point are also output. The obtained laser mirror system has ideal effect.

The methods of synthetic aperture radar (SAR) jamming effect evaluation is a current research hotspot in the field of electronic countermeasures. Aiming at the problem that the existing SAR jamming evaluation techniques are only applicable to cooperative jamming effect evaluation and cannot provide real-time guidance to the jamming strategy of jammers on the battlefield. Our paper proposes a model which can predict the effect of SAR noise barrage jamming based on the image subjective evaluation standard—the National Image Interpretability Rating Scale (NIIRS). Two factors with the most significant impact on NIIRS, SAR image resolution and jamming-to-signal ratio, are selected as the independent variables of the model. This model not only enables the prediction of jamming effect, but also reflects the jamming effect to the loss of image interpretability directly. Experimentally, this model is fitted with a large amount of SAR image data, and the R2 is 0.91, RMSE is 0.41.

This study designs an algorithm to reconstruct the marker point cloud using a monocular camera. The specific processes can be described as: Firstly, a group of images containing only the coded marker and the reference ruler are used for selfcalibration; Secondly, another group of object images are reconstructed by the internal parameters of the camera; The processing algorithms of the two groups of images are similar, and the second group needs additional matching and reconstruction of noncoding points. When a large number of coding marks cannot be placed on the surface of the object to be measured, the observed value of adjustment is reduced. The experimental verification of industrial structural parts shows that the measurement error is 0.05 mm and the relative error is 0.02% in the measurement range of 5 m. Compared with the method using camera internal parameter as observation value adjustment, the accuracy is similar, the number of observations is reduced and it can be applied to most industrial measurement fields.

In this paper, the DJI Phantom4 RTK unmanned aerial vehicle is used to monitor the changes of the beach in the northeast of Zhoushan Island in two stages, and the DSM (Digital Surface Model) and DOM (Digital Orthophoto Map) are generated. RTK-GNSS elevation point acquisition is used to ensure the accuracy of the model. After the model elevation is revised, the mapping accuracy is evaluated and tested by error formula and linear fitting method. At the same time, the geological interpretation and profile elevation monitoring of the research beach in Zhoushan Island were carried out by using the two-stage high-precision model. The results show that the surface micro-geomorphology has disappeared and the nature of the beach has changed. The aeolian dune in the front shore increases by 0.5 m, and the overall elevation of the beach surface (intertidal zone) in the front shore area decreases by 0.4 – 0.6 m, showing the phenomenon of erosion in the front shore and deposition in the back shore.

The 2022 Winter Olympics will be held across three zones: Beijing, Yanqing and Zhangjiakou. Zhangjiakou will host the majority of ski and snowboarding events. Therefore, the analysis of spatiotemporal change in Zhangjiakou is of great significance to the preparation and sporting event host of the Winter Olympics. This study takes Zhangjiakou as the region of research, bases on the Google earth engine. MODIS Normalized Difference Snow Index dataset ranging from 2000 to 2021 is used for extracting snow cover data in the study area, then comparing with the calculated snow cover by Landsat 8 Surface reflectance data, as well as combining meteorological observation data. The spatiotemporal variation characteristics of snow cover are analysed. The result shows that there is a good agreement in snow cover data between MODIS and Landsat 8; the MODIS data can reflect the distribution characteristics of snow cover area; the area of snow cover and days of yearly snow lying show periodic fluctuation and trend of weak growth. The distribution of snow cover in Zhangjiakou shows a ribbon pattern and planar orientation. The frequency of snow cover presence in the whole area is mostly less than 0.4 while the frequency in the area with sporting venues is mainly more than 0.6, and remains around 0.5 in February when the Winter Olympics is holding.

Remote sensing can be used to monitor and warn of the damage of external force caused by human activities such as road construction and building construction along the underground cable. The extraction of external force damage information is mainly based on the results of change detection. Based on the object-oriented change detection method, our study selects the optimal features through multi-scale segmentation, multi-source feature construction and classifies the two-time high-resolution satellite remote sensing images. Using the classification results to establish the hierarchical relationship, the idea of superposition analysis is introduced, and the detection results of ground object changes are obtained by logical difference operation, then the external rupture information is extracted. This method avoids the subjectivity and complexity of the varying threshold method. Our results show that the detection probability is 88.63%, the false detection probability is 11.36%, and the miss detection probability is 4.41%.

Road extraction from remote sensing image is a fundamental task. Although, the methods based on CNNs have achieved great progress. It is difficult for network-based on CNNs to achieve a breakthrough in performance due to the limitation of receptive field. However, Transformer has better capabilities to build the global receptive field than CNNs. This paper proposes a novel network called TransLinkNet which combines CNNs and Transformer to obtain robust feature representations. Specifically, a stack of Transformer blocks is interspersed between LinkNet layers. Convolution operations are good at obtaining local features, while the attention mechanism in Transformer to build the global receptive field. Experiments have proved that the model achieves competitive performance on The Massachusetts dataset.

A portable laser induced fluorescence (LIF) lidar is designed and developed for oil pollution monitoring. A 405nm violet light is used as excitation light source. A dual Amici prism spectroscopic structure to split the fluorescence signal in the spectral range of 450-1000nm, and an image intensifier combined with array CCD to achieve high sensitivity detection. The system integrates control, data acquisition and data processing modules, which can realize the real-time detection of the composition and content of pollutants. The fluorescence spectra of seawater, diesel oil, oil and crude oil were measured at 30m. The spectra of diesel oil, engine oil and crude oil with concentration at 0.05, 0.01 and 0.01mg/L are measured. The LIF oil pollution monitoring system can accurately distinguish oil spill pollutants such as type and grade. It has small volume, light weight, high integration, high detection sensitivity and good stability characteristics, which can be loaded in small aircraft, unmanned aerial vehicles, ships for oil spill remote sensing.

Taking into account the complexity of the terrain, the triangulation irregular net thinning method is used to obtain the characteristic spot of multi-beam sounding data, and the advantages of multi-diagonal exchange algorithm are integrated and improved, and the improved constraint algorithm is used to embed the isobaths to construct the seabed DEM. Experiments indicates that the method can effectively identify the characteristic spot, and the thinning method based on characteristic spot and isobath constraint can effectively construct the seabed DEM with high precision and high resolution.

Vegetation is an important part of any ecosystem, and its change has a far-reaching impact on regional and even global ecosystems.Normalized vegetation index (NDVI) is often used to reflect the growth status of vegetation. Combined with the seasonal change trend of topography and vegetation in the region, the evolution characteristics of vegetation temporal and spatial pattern in the region are analyzed. This paper selects Chengdu as the research object. Through the processing of Landsat8 remote sensing data in the third quarter of each year from 2013 to 2020, combined with Chengdu administrative regional data and DEM data, the temporal and spatial pattern change law of normalized vegetation index (NDVI) in Chengdu from 2013 to 2020 is studied by using trend analysis method and stability analysis method. According to the above analysis, taking all district level administrative regions of Chengdu as the center, the vegetation in Chengdu shows a fluctuating growth of low outside and high inside, and the vegetation outside the urban area tends to grow steadily; Spatially, the vegetation in the area below 2000m increases gradually with the increase of altitude, while the vegetation in the area above 2000m decreases gradually with the increase of altitude.

Aiming at the problems of large amount of laser point cloud data, uneven distribution of point cloud and unsmooth surface of reconstructed model, a moving least square method with value-added conditions is proposed to smooth sample the point cloud, and by setting the search neighborhood radius, the point cloud surface is smoother and the detail features are clearer; At the same time, a Delaunay triangular mesh reconstruction algorithm based on multi criteria is proposed to realize the model reconstruction of point cloud after smooth sampling, so as to improve the reconstruction accuracy of large-scale point cloud, and improve the effect of meshing model by manually adjusting parameters. The feasibility of this method is verified by comparative experiments.

Map projection, benefiting from its visualization analysis, becomes more intuitive and reasonable. It has significant applications in teaching activities, scientific research, and many other fields. In the absence of systematic elaboration of visualization analysis methods in modern map projection field, theoretical analysis of independent map visualization can effectively solve the difficulties in systematic visualization work. Based on the existing map projection theory, this paper researches the visualization platforms, methods, and application mode. The main work of this paper includes expounding essential theoretical category of projection visualization and systematically discussing the visualization method of projection analysis. Finally, the projection deformation visualization analysis is carried out and applied in the projection selection of the Arctic.

The earth's gravitational field is one of the basic physical fields of the earth, which reflects the spatial distribution, movement and change of the earth's internal matter, and is the external expression of the earth's material distribution and the earth's rotation movement information. As one of the main objectives of modern geodesy, high precision geophysical field information measurement is also an important support to establish global gravity field model and determine geoid. As the main method of high precision gravity survey, sea-air gravity survey has the characteristics of high speed, high efficiency and high precision. The process of sea-air gravimetry is mainly divided into gravity data acquisition and gravity data processing, and as the transition between the two, the calculation of the intersection of the sea-air gravimetry line network plays an important role. In view of the low precision and complex operation of the current calculation method of intersection search, this paper proposes a new method to realize intersection search based on the actual intersection of survey line network, and verifies the simplicity and effectiveness of this method through simulation data and measured data. The results show that the accuracy of the new method is about 0.02 mGal higher than that of the slope method of primary and secondary survey line, and the calculation speed is faster than that of the determinant method.

Because of the continuous advancement of spectral imaging technology, the spectral fineness of hyperspectral remote sensing images increases continually, and the scale of band data enlarges constantly, it is more and more difficult to process multi-dimensional data. Band selection has been applied in many fields as an effective way of data dimensionality reduction, however, the relatively single experimental data used by the existing common methods, the ability to discriminate when applied to specific objects such as camouflage targets cannot be verified. Therefore, when the research uses the band selection method to solve the problem of large amount of data and band redundancy in hyperspectral images, how to quantitatively calculate the significance of the camouflage target has become an urgent problem. The image evaluation index of simulated vision is introduced, and the quantitative results are used to show the effectiveness of the selected method for identifying the target, which provides theoretical support and thinking reference for the establishment of an algorithm model suitable for the recognition of camouflage targets in the battlefield environment.

Interferometric radiometers play an important role in microwave remote sensing. The measurements provide massive amounts of information for earth observation. However, Radio Frequency Interference (RFI) sources can deteriorate scientific data quality, if they haven’t been detected properly. In this paper, we describe a RFI detecting method based on multiple statistical characteristics, and experiments have been carried out on global brightness temperature to demonstrate its effectiveness.

By using the computer algebra system, the formulas of Mercator and transverse Mercator projection are weighted and combined according to certain proportional coefficients, and a new combined projection method is obtained. Taking China as an example, the influence of different scale coefficients on combined projection deformation is investigated. Through comparative analysis, the weighted combination coefficient suitable for China region with minimum projection deformation is obtained. The combined projection not only keeps the original equiangular property, but also reduces the projection deformation to a certain extent.

Taking Lanzhou Botanical Garden as the study area, the visible light remote sensing images in the study area from March 1, 2021 to June 7, 2021 were gathered by using a drone (Phantom 4 RTK). The Excess green index (ExG) was extracted and the inversion models between this index and fractional vegetation cover (FVC) were analyzed to explore the optimal inversion model in the study area. This index was used to establish the FVC inversion model in the study area. We compared the regression model results of polynomial, exponential, logarithmic and others, and the optimal FVC inversion model in the study area was the model of function of first degree: y=0.37x-1.67. It could estimate the dynamic change of FVC in the study area. Through this study, we can provide a scientific basis for the protection of the ecosystem in Lanzhou and other semi-arid area.

The height (H) and above-ground biomass (AGB), as important phenotypic parameters, are important guides for growth monitoring and genetic improvement of winter wheat. In this study, digital orthophoto map (DOM) and digital surface model (DSM) of winter wheat at the filling stage were obtained based on different water and nitrogen treatments, using a UAV-mounted multispectral imager. Then a model was established for estimating height and biomass using the partial least squares method. The results showed that (1) at the same nitrogen application level, when each sample in various nitrogen levels was modeled separately, the model for estimating height established by N0 and N6 was better, with R² of 0.773 and 0.592, respectively; when all samples were modeled together, the R² was 0.831, which was better than the former. (2) The model of the relationship between height and measured biomass extracted by UAV under different nitrogen levels was robust, and the estimated model R² and RMSE were 0.598 and 4308.99 kg/ha, and the validated model R² and RMSE were 0.507 and 4939.31 kg/ha. (3) The R² and RMSE of the model using biomass per unit plant height with measured biomass were 0.881 and 2335.46 kg/ha, and the R² and RMSE of the transformed biomass estimation model were 0.876 and 2425.39 kg/ha, which improved the accuracy by 72.7% compared with the estimation model built by directly extracting the height using UAV. This indicates that DSM can accurately estimate the height and biomass of winter wheat, providing a reference for phenotypic studies as well as water and nitrogen management of winter wheat.

We demonstrate a reliable network to disseminate high-precision RF frequency by optical frequency comb via fiber link. Dense wavelength division multiplexing and MEMS optical switches are used in this network, which realize protection switching. In this network, high-precision frequency is transferred to multi-nodes. The fiber link is monitored in real time and the synchronized signal can be automatically recovered immediately. We propose a phase noise compensation technique that the fractional frequency instability was measured to be at 9.2 X 10^-17/10²s on 20 km optical fiber link. This work paves the way for ultra-stable frequency transmission and distribution networks.

Aiming at the problems of misdetection and missed detection in traffic signal detection tasks in current traffic scenes, this paper proposes an improved YOLO v3 traffic signal recognition algorithm YOLO v3-SimAM. First, introduce the new attention mechanism SimAM to enhance the features; secondly, add the SSH network structure after the feature layer extracted by the FPN feature pyramid to expand the deep network receptive field to further strengthen the feature extraction; finally, the last convolution of YOLO Head is replaced with dynamic convolution to improve the accuracy of target frame detection. At the same time, in the loss function design, the original intersection ratio (IoU) loss function is replaced with the DIoU loss function, and finally the YOLO v3-SimAM algorithm is formed. The algorithm is applied to traffic signal detection tasks. On the public TTTL (Tsinghua-Tencent Traffic Light) data set, the average accuracy of the algorithm (mAP) index reached 67.67%, which is an increase of 1.2% compared with the YOLO v3 algorithm. The results show that compared with the original YOLO v3 algorithm, the YOLO v3-SimAM algorithm proposed in this paper can detect traffic lights more accurately.

As two core fields of machine learning, Natural Language Processing and Computer Vision have derived a variety of cross research topics, and image captioning as a key topic of the current cross field has an important relationship with the development of both. Image captioning has been attached importance to by researchers for a long time, and has formed a relatively perfect theoretical model system. At present, this topic has made great progress in accuracy, but this is only one aspect of the evaluation of image content description algorithm, which has new requirements in diversity, stylization and controllability. Based on this, this paper studies the technical routes and modeling methods of various auto-coding models, variational models and attention mechanisms, proposing a new way to solve the inherent contradiction between diversity and accuracy, that is to increase the diversity of each sample while ensuring the upper bound of accuracy index scores in multiple samples. The test results on MS COCO data set show that the model implemented in this paper has good performance in accuracy and diversity.

In recent years, sustainable development has attracted more and more attention, and waste classification is a social problem related to people's livelihood and social sustainable development. Therefore, this paper proposes a waste classification system based on deep learning. The system is built with tensorflow framework in deep learning and data training with mobilenet model, The data set is obtained by the method of web crawler, cleaned and classified, and the graphical interface is constructed by pyqt5. Finally, by uploading the pictures, the system can effectively classify the objects in the pictures into the correct garbage name and category.

Accurate and rapid measurement of the notch of vehicle accessories produced in real time on the assembly line is an important way to reduce labor cost and improve production efficiency. The current dimension measurement algorithms pay attention to the extraction of edge high-precision contour which needs complex parameter adjustment and despise the extraction of contour corner information. It is difficult to be used in industrial real-time detection environment because of its large amount of calculation, easy to be disturbed by the environment and weak robustness. In order to realize the noncontact, high-precision and wide application range of notch length measurement based on machine vision, an algorithm is proposed to extract corner points with contour point features and contour edges with geometric restrictions, so as to finally obtain the length of notch. The algorithm uses moments to estimate the centroid coordinates and overall direction of the accessory, and calculates the distance to the centroid and the offset angle from the overall direction for each side of the accessory one by one. The results show that compared with the traditional corner extraction and line detection algorithms, the algorithm can extract the accessory corner and the notch length more accurately. The average processing time for the accessory image with a resolution of 200 pixels (vertical) × 600 pixels (horizontal) is 15 milliseconds, and the notch length can still be extracted for the slight shaking generated in the manufacturing process. The relative error of the extracted notch length is less than 1%, which is suitable for the industrial field environment with high requirements for real-time and robustness. It has the characteristics of strong adaptability and low cost.

Aiming at the problems of image super-resolution reconstruction method based on convolutional neural network, such as complex structure, huge parameter amount, and slow reconstruction speed, this paper proposes a multi-attention feature fusion network for lightweight image super-resolution. The channel attention mechanism and pixel attention mechanism are used to fully extract image feature information and improve the feature extraction ability of the network. At the same time, the use of depthwise convolution effectively reduces the amount of network parameters and calculations. The experimental results show that under the premise that the reconstruction performance of the network is competitive, the network is lighter and the reconstruction speed is further improved.

Aiming at the problem of the classic two-stage object detection algorithm Faster R-CNN that it is difficult to fully extract and fuse multi-scale features, an improved Faster R-CNN algorithm based on switchable atrous convolution and involution is proposed. This method first replaces the 3×3 convolution in the feature extraction module with a switchable atrous convolution to better capture the original rich information of the feature; then, replaces the 3×3 convolution in the feature fusion module with an involution further enhances the feature fusion effect at different stages. In order to test the detection effect of the algorithm, experiments were carried out on the MS COCO data set and the PASCAL VOC data set. Compared with the improved Faster R-CNN algorithm, the average precision of the proposed algorithm on the two data sets is improved by 5.5% and 2.2%, respectively. Among them, for the detection of larger objects in the COCO data set, the effect is improved more significantly.

Recently, deep learning has been rapidly developed in the field of object detection. However, how to accurately and efficiently locate rotating objects is still a challenging task. This work adopts the YOLOv5 algorithm as a rotation detector to locate the rotated objects in images. We first analyze and discuss the angle parameter regression method now commonly used in angle prediction, the angle classification method and the circular smooth label rotation detection method. Then, we validate these methods in the widely recognized rotation target image DOTA dataset. The experiments show that the average accuracy of the three methods is about 72% on rotation objects detection, and among these three methods, the circular smooth label rotation detection method achieves the highest performance on the task of rotation prediction.

In this paper, we propose a semantic-aware network (SA-Net) to improve the performance of 3D point cloud object detection, which embeds a backward attention module and a semantic attention module. The backward attention module utilizes high-level semantic features from the encoder via fusing multi-level encoder features hierarchically. In this stage, high-level features are transformed into an attention map to modulate low-level features backward. Meanwhile, semantic attention module obtains a semantic segmentation map of a given point cloud scene through supervised learning. This can be transformed into a semantic attention map and embedded into the detection head for better detection. Equipped with these modules, SA-Net can greatly improve the performance of object detection. Extensive experiments on KITTI demonstrate that the proposed method can achieve competitive results against the state-of-the-art methods.

Aiming at the characteristics of complicated partial discharge test equipment and strict operating procedures for insulating parts, a set of intelligent interactive system for partial discharge tests of insulating parts based on AR technology has been developed. This article first sorts out the operating procedures of the test equipment, and analyzes the data and scene interaction mechanism during the test. Then explore the use of 3D laser scanning technology, space recognition technology, simultaneous localization and mapping (SLAM) technology, AR technology to build a system technology framework that includes the underlying data layer, technical support layer, user layer, and application presentation layer. Finally, the test results of the navigation function in the application presentation layer are shown. Through on-site testing, HoloLens AR glasses can realize the intelligent perception of the electrical test environment and key parts of the test equipment after loading the intelligent interactive system, achieve the effect of intelligent interaction, and achieve the purpose of improving test efficiency and ensuring operation quality and safety.

Visual symbols are the main component of visual design. Professor Wei Huang proposed the eye-opening theory of visual symbols in 1995 and pointed out the effect of changes to the whole. Based on the principle of visual closure in Gestalt psychology, for a figure with missing part, the human perceptual impressions will subject it to subjective processing, either to complement the missing part or to separate it from the whole. However, what is the most noticeable missing ratio that can cause visual retention? This paper will use the classic point detection paradigm of psychology to design experiments to measure the eye-opening perception of visual symbols.

With the rapid development of urban traffic, the problem of urban traffic becomes very serious. The traditional signal lamp control method can not adaptively control the traffic signal at the intersection. The rise of reinforcement learning technologies has made traffic control and artificial intelligence closely related. This paper presents a traffic optimization algorithm and model based on SUMO simulation platform. In this model, SARSA algorithm in reinforcement learning is used to establish multi-intersection simulation model. The multi-intersection simulation model is four consecutive intersections in Guangming Road, and the experimental parameters are actually investigated. By comparing the simulation data of multiple intersections, it is found that SARSA algorithm is superior to traditional fixed timing and full induction control.

Covid-19 pandemic continues to threaten health of the global population, an efficient way to restrain the Covid-19 outbreak is timely screening suspected cases for quarantine and treatment. Despite of pathogenic laboratory testing is the gold standard to screen suspected cases, but it may obtain false negative results and consuming a lot of time. Computed tomography of chest can be an alternative diagnostic method to screen suspected cases that is based on radio graphical changes in lung area of Covid-19 confirmed case. Precisely delineate the lung area that is first and critical step for screening computed tomography image of chest by using deep learning method. In this paper, several related previous works will be introduced at first. Then, an improved encoder-decoder based segmentation framework is proposed, which is integrated with multi-scale densely connection-based convolution block and skip connection path. Moreover, in model training process, a semi-supervised manner is applied to train model which can reduce the demand of labeled training data. Finally, the proposed method is tested with a dataset of public X-ray image of chest. The experiment test proposed model in this paper with varieties of segmentation methods and result demonstrates promising performance of proposed model that against several other deep learning models.

With the rapid development of artificial intelligence, reinforcement learning techniques are emerging, and in the field of transportation, the control of traffic signals can be optimized using reinforcement learning algorithm to improve traffic efficiency. In teaching, the reinforcement learning algorithm is more difficult than the traditional timing algorithm to understand, and there is a lack of teaching platform - which can not only give students an in-depth understanding of the algorithm and advantages of reinforcement learning, but also facilitate teachers to evaluate students' mastery. In view of the above problems, this project designs and develops a traffic signal control optimization simulation teaching platform using AI algorithm, which realizes a number of reinforcement learning algorithms, and combines the basic characteristics of teaching, adopts a distributed system structure, which can meet the needs of online teaching and evaluation.

The platform is divided into teacher side and student side, the main functions of the student side are: SUMO road network simulation model selection, action mechanism selection (and input related parameters), reinforcement learning algorithm selection (Q-learning and Sarsa), model input parameters and evaluation parameters (such as fuel consumption, queue number) selection, algorithm operation, statistical output and so on. During the operation, the platform displays the statistical results in real time in the form of dynamic graphs, and after the operation, outputs static charts of evaluation data of different algorithms. The platform has the function of evaluating and scoring according to the weighted average of the optimization of output results. According to the requirements set by teachers, it can score the aspects of pass efficiency and environmental impact, and generate experimental reports as the basis for teaching evaluation. The teacher can view the completion of the experiment and the evaluation results of the experiment, so as to check the students' learning of the reinforcement learning algorithm.

The work done in this subject, on the one hand, helps students to study and experiment with advanced AI algorithms, on the other hand, it can also help teachers to master teaching situation, improve teaching quality, and it is of great significance to the application of AI algorithms in the field of traffic signal control optimization and personnel training.

With the rapid development of communication technology and social media, people are constantly interacting with each other, generating a large amount of behavioral data and forming a complex social network. With time, the interaction behaviors between people also change, which results in a constantly changing social network. Analysis of evolutionary social networks can effectively mining people's behavioral data, understanding people's behavior models in depth. Community detection is one of the important methods for social network analysis, and evolutionary analysis of dynamically changing communities, especially the core stable community structure in dynamic networks, can reveal the community structure at different periods and reveal the social network evolution patterns. In this paper, we define communities on real complex networks as core-edge structures with a tight core and sparse periphery, which can represent the importance of nodes within the community. And we propose a community detection algorithm based on the core-edge structure communities. Moreover, based on the characteristics of the core-edge community structure, we propose a community similarity calculation method based on the importance of members and an evolution analysis method based on the core-edge community structure.

Traditional vehicle target recognition technology mainly uses artificial or artificial intelligence methods for recognition in the visible light band. Aiming at the problems of low level of intelligence and low efficiency of vehicle recognition at night and low visibility, an improved SVM recognition algorithm based on infrared images is proposed. In this research, the infrared images of small vehicles, medium vehicles and large vehicles at night are used as the object, and the infrared characteristic information of vehicles is extracted based on the histogram of directional gradient (HOG), and the internal parameters are optimized by improving the support vector machine. After the algorithm is trained, the accurate recognition rate of all vehicles can reach 98%, and the effect of the improved algorithm is significantly better than that of the unimproved algorithm, which provides a certain reference for the application of machine learning in vehicle target recognition.

In order to realize the test for equipment system, shorten the test period, reduce the test consumption and speed up the process of equipment assigning to users, the joint implementation method of multi-type equipment is proposed. Firstly, the relevance of the joint implementation with different single-type equipment is analyzed, including the joint implementation correlation analysis and the independence analysis of test evaluation; Then the construction method of system capability index and single-type equipment index is presented; Finally the model based on subjective weight, objective weight and combination weight is expounded respectively.

Aiming at the current situation that some drivers violate the regulations and overspeed when driving in multi-type tunnel sections, a tunnel speed limiting device based on probability model and license plate recognition is proposed to monitor the overspeed of vehicles in tunnel sections. Firstly, a camera is arranged at the entrance and exit of the tunnel to capture the license plate images of vehicles in and out. The license plate images are preprocessed by improved gray transformation, and the images are enhanced according to the tunnel structure and environmental characteristics. The Prewitt operator is used to detect the edge of the preprocessed images. The oTSU threshold segmentation algorithm is used to segment the target region from the image, and the probability model is established based on the parameters of gray co-occurrence matrix, and the feature region is screened to recognize the license plate. The time was recorded when the vehicle entered and left the tunnel, and the average speed was compared with the speed limit value to judge whether the vehicle was overspeed. MATLAB software is used to simulate the experimental process, and the test proves that the device can accurately judge the overspeed situation of vehicles in the tunnel, which is of certain significance to the tunnel speed limit monitoring.

The multi-state rapid recognition of switches is the basic service of the AR intelligent interactive system. In order to minimize the computational burden, this article first analyzes the state characteristics of different types of switches, and introduces the advantages of HoloLens2 AR smart glasses in the recognition of the switch state. Then, according to the degree of importance, the laser point cloud scanning technology is used to generate a three-dimensional model of the equipment object and the surrounding environment that takes into account the accuracy and the amount of data. Finally, the visual SLAM algorithm technology based on the feature point method is used to complete the rapid recognition of the switch multi-state. Practical applications show that, for the switch state on the device, the recognition rate of the AR technology-based switch multi-state rapid recognition method proposed in this paper is up to 100%, which meets the needs of the AR intelligent interactive system.

With the rapid development of modern inspection technology, manual rice quality inspection technology has fallen behind, and rice quality inspection technology based on machine vision has become inevitable. However, the current research results and testing technology cannot fully meet the testing standards of the national rice industry, and a more rapid and accurate quality testing system is needed. This paper uses machine vision technology to carry out related research on the application basis of rice quality inspection, and gradually studies the problems of chalkiness detection and gel consistency detection encountered in the process of rice quality inspection, and lays the foundation for accelerating the construction of a rice quality inspection system based on machine vision.

With the development of the "digital" era, film reviews have moved from words to numbers. General film ratings use expert or public ratings as a single type of evaluation index to score films according to different weights. In this paper, different types of evaluation indexes are selected in a multi-dimensional way. The evaluation indexes are divided into pre-screening period, screening period and post-screening period in the time dimension. The most representative evaluation indexes in different periods are selected for multi-dimensional evaluation of film quality in this period. Finally, the overall score of the whole film is combined with the score of each period. The weights of each evaluation index will be different according to their own characteristics, and the weights of this paper are basically determined by analytic hierarchy process (AHP). Scores for each period were calculated based on Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) algorithm. Based on the analysis of data, not only consider a fixed value, at the same time consider over time due to the evaluation, heat and other factors on the impact of film data, through the establishment of basic data rate and deeper rate data processing model of change over time, more accurate data for processing, then makes the movie score is more accurate.

Trajectory tracking technology is one of the key technologies of unmanned vehicles, and accurate trajectory tracking control is also the primary premise for unmanned vehicles to complete various military tasks. At present, the trajectory tracking research of unmanned vehicle mostly relies on the research results of mobile robot. The traditional trajectory tracking control methods mainly include PID control, neural network control, optimal control, sliding mode control and predictive control. They have their own advantages and limitations. By studying the control behavior of vehicle drivers, using intelligent control method, this paper establishes a nonlinear time-varying dynamic model suitable for unmanned vehicles, breaks through the key technologies such as accurate trajectory tracking control of unmanned vehicles, and provides technical support for the development of new unmanned vehicles in practical environment and the unmanned transformation of active vehicles.

For the problem of low-illumination underwater images caused by the use of filters, we provide an algorithm for image enhancement based on multiscale Retinex and Dark channel priori. First, we use the multi-scale Retinex algorithm for color correction. Second, we use a modified Dark channel priori algorithm for the defogging process. We acquired some underwater images and compared them with classical image processing algorithms, and the results show that our algorithm improves image brightness and contrast.

Corner detection is a method used to obtain image features in computer vision systems, and is widely used in motion detection, image matching, video tracking, three-dimensional modeling, and target recognition. Also called feature point detection. Aiming at the shortcomings of traditional fast corner point extraction algorithms that use fixed thresholds when extracting image feature points and are easily affected by illumination to extract external points on a smooth plane, this paper proposes an adaptive threshold FAST corner detection method combined with edge information. First, use the canny algorithm to extract the edge information of the image, and then explain the calculation method of the adaptive threshold, calculate the feature point extraction threshold according to the gray information of the image, and finally extract the FAST corner points from the edge information according to the adaptive threshold. The test results show that: compared with the original FAST corner detection algorithm, the improved algorithm has increased its running speed by more than 30%, and has significantly reduced the number of external points, and its noise resistance has also been significantly improved.

The Transformer structures have achieved excellent performances in an increasing number of fields. In the field of natural language processing, the original Transformer structure was used to work on a full connectivity graph, which depicts all connections between words in a phrase. It has been subsequently applied in areas such as computer vision. Recently, many studies have attempted to apply Transformer structures to graph data and achieved good performance. However, it is still a challenge to embedding the structural information of the graph data. To understand the representation of graph data, we employed three distinct ways of embedding the positional and structural information of graph data based on the Transformer structure in this work. To learn the feature representation of the graph data, the encoded structural information and original feature information of the graph data were fed into the Transformer structure. For the learned feature representations, we used them for node classification and graph regression, and then compared and analyzed the three different encoding methods.

Terrain texture attributes are an important means to characterize different types of landform units. The combination of geomorphic units reflects the spatial distribution pattern of regional geomorphic features. Therefore, the analysis of terrain texture significance and variability is a key link in the study of regional geomorphological features. In the process of extracting and analyzing regional geomorphological features, traditional analysis methods of macroscopic and microscopic morphological indicators rely on window areas and calculation models. Although the morphological feature element analysis method can reveal the basic pattern of regional geomorphology, it is aimed at the spatial change structure of terrain elements. In order to grasp the overall and local hierarchical structure characteristics of regional geomorphology, we propose a comprehensive metric model of terrain texture based on texture feature entropy weight method in this paper. Nine single geomorphological research areas were selected to carry out experimental research in Shaanxi Province, China. The results show that the regional characteristics mapped by the comprehensive metric model of terrain texture show the similarity of similar landforms and the difference of different landforms at a certain scale. The closer the metric model index value is to 1, the more significant the comprehensive feature of texture is, the stronger the periodicity of landform is, and the more uneven the spatial distribution of texture is. This model is helpful to promote the overall and systematic research on digital terrain analysis of regional geomorphology in China.

The traditional surveying and mapping of ancient landslide is mainly based on GPS, which has high precision but low efficiency. As a new remote sensing technology, synthetic aperture radar interferometry (InSAR) has developed into one of the effective means to quickly obtain high-precision digital elevation model. In this paper, the shenjiazui ancient landslide formed by the Haiyuan earthquake in 1920 is taken as the research object. The elevation information of the ancient landslide is obtained by three means: field RTK measurement, DEM extraction based on Sentinel-1A radar data and SRTM download. The DEM accuracy is compared by three methods: checkpoint method, profile method and contour playback method. Through the precision comparison and analysis, it is found that the DEM extracted by Sentinel-1A is in good agreement with the DEM measured by field RTK. The mean error is within 0.6m and the mean square error is within 9m, which is much better than the currently disclosed SRTM precision. It can quickly, widely and accurately obtain the surface 3D data and apply it to the 3D modeling of ancient landslide.

As an important element of a smart city, building outlines are widely used in the fields of urban space analysis, building classification statistics and digital city modeling. Ground laser scanning captures three-dimensional point clouds with high flexibility and high precision, and provides important data support for extracting building outlines. The traditional method of extracting building contour is easily interfered by other factors, which leads to the problem of jagged contour extraction and low extraction accuracy. The variable radius A-shapes building point cloud contour extraction method based on the least squares proposed in this paper combines the advantages of least squares straight line fitting and rolling circle detection to optimize the key contour points, which can effectively remove the jagged contour lines. It has high extraction accuracy and good robustness. The experimental results show that, compared with the Variable radius Alpha Shapes (VA-Shapes) algorithm, the building outline extracted by the method in this paper is clear and complete, and the extraction effect of the building outline with more corner features is better. The accuracy of the obtained roof contour line is higher.

Object detection is a research hot issue. With the improvement of remote sensing technology, the task of remote sensing target detection has become more important. Features of remote sensing aircraft images: small and dense targets to be measured. Common small object detection methods are implemented using CNN. In the small object detection task, since its convolution structure, information is lost. The Transformer's non-convolution structure can increase the receptive field while making full use of the image feature information, but because its attention calculation is between the tokens, the internal information of the token cannot be calculated. Therefore, we improve the encoder part and construct a scale fusion module to solve this problem. Experiments prove that the network proposed in the paper has good performance on the DIOR dataset.

It is of great significance to study the evolution characteristics of the spatial-temporal pattern of urban air quality and its influencing factors for deeply understanding the mutual feedback mechanism between urban environment and socioeconomic system and formulating efficient environmental control measures. Based on the real-time monitoring data of PM2.5 in Shaanxi Province in 2015, this paper discusses the temporal and spatial distribution law of Shaanxi Province, and uses the random forest model to reveal the nonlinear driving mechanism of meteorological factors, topography, vegetation, human activities, time and space factors on PM2.5, and sort out the key influencing factors.The results showed that the concentration of PM2.5 in Shaanxi Province not only showed a "U"-shaped monthly variation law, but also showed the characteristics of high in winter, followed by spring and autumn, and low in summer, with obvious seasonal differences. In space, it generally presents the distribution characteristics of northern Shaanxi < southern Shaanxi < Guanzhong. The model shows that aerosol optical depth (AOD) and boundary layer height (BLH) have relatively significant effects on PM2.5 concentration, while precipitation has the least effect on it.

Due to the coherence of the scattering phenomenon, synthetic aperture radar (SAR) images are contaminated by multiplicative speckle noise, which seriously affects the interpretation and identification of SAR images. In this paper, we introduce the curvature filters which are realized by constructing tangent planes and correcting the regularization energy into speckle noise suppression of SAR images. Compared with the ROF model, the guided filter and some classic SAR image filtering algorithms (such as Lee filter), the curvature filters show strong competitiveness in the computing time, the peak signal-to-noise ratio (PSNR) and the structure similarity (SSIM) index. In addition, we compared the performance of the three curvature optimization models of the curvature filters, and found that the total variational filter model performs better in dealing with speckle noise of SAR images, which shows that the construction of local half-window and the assumption of piecewise constant for SAR images are robust to speckle noise suppression and structure preservation.

Aiming at the problems of single considerations, poor universality, and strong subjectivity in the semantic similarity calculation method for the geographic information field, combined with the structural characteristics of the geo-ontology, an algorithm for semantic similarity of geo-ontology concept based on BP neural network is proposed. Based on the traditional calculation of semantic distance and information factor based on the ontology hierarchical tree structure, the ontology properties value and semantic relationship between geographical concepts are extracted based on the authoritative concept description in the field, and they are transformed into the expression of semantic similarity. Synthesize the influence of four types of factors on semantic similarity, and use BP neural network to optimize the weight. Calculating the correlation coefficient between this algorithm and the expert evaluation results and comparing it with other algorithms shows that this method can effectively improve the accuracy and effectiveness of the semantic similarity calculation of geographic information concepts.

Due to the fixed length and short distance of the existing length reference ruler in the digital industrial photogrammetry system, it is difficult to meet the measurement of large-size components. This article proposes a variable-length reference ruler construction method. Firstly, a tooling suitable for total station and photogrammetry is developed, and the tooling is placed at the two ends of the largest dimension of the component to be tested. The maximum actual length of the component is measured by the total station, the point coordinates with the non-real size length of the component is obtained by photogrammetry. Then, the three-dimensional coordinates of all points with the actual scale factor are obtained, by bringing the actual size measured by the total station into the photogrammetric calculation model, and the calibrated size of the tooling itself can be used to verify the accuracy of the length reference measurement of this method. At last, it is verified through examples that the variable-length reference ruler construction method of this article overcomes the shortcomings of the existing photogrammetric system’s own length reference ruler, it provides an effective way to carry out convenient and efficient high-precision measurement of large-size components, and improves the on-site adaptability of the photogrammetric system.

The GF-7 satellite has high spatial resolution. This article compares GF-7, GF-1 and GF-2 in terms of image color effect, spectral resolution and feature difference, and makes a preliminary evaluation of the accuracy of remote sensing geological structure interpretation and lithology information recognition based on GF-7 data. The results show that GF-7 can accurately identify geological bodies, and it can be used to carry out remote sensing geological information extraction from 1:5,000 to 1:20,000.

To provide flexible and efficient geographic information services, a new asymptotic transmission method based on a linear binary line generalization (BLG) tree storage structure was proposed. First, we used a BLG tree to store all the hierarchical simplification results of the dynamic programming (DP) algorithm and modify the sub-node offsets. Then, we set different target scales and calculated the average distance between neighboring feature points to filter out the appropriate feature points. Finally, the protein inference algorithm (PIA) iteratively approximated the set using three non-uniform B-sample curves to improve the accuracy. The experiment results proved that the proposed could effectively improve the network transmission efficiency of vector map data.

Global Navigation Satellite System (GNSS) technology provides an effective means to invert the Zenith tropospheric delay (ZTD), which can be obtained with high accuracy and high temporal resolution independent of weather variability. As a key parameter in climate change studies, ZTD is widely used in El Niño-Southern Oscillation (ENSO), air quality, and drought and flood monitoring studies. However, due to the sensor or observation equipment, some ZTD data are wrong or missing, which leads to biased analysis results of related studies. Therefore, in this paper, the daily-scale ZTD of 40 GNSS stations in the Chinese region from 2011 to 2020 is used as experimental data, and the K-Nearest Neighbor (KNN) and Periodic Model (PM) methods are used to complete the data and evaluate the accuracy considering the missing rate of 10%, 20% and 30%, where The KNN method takes the number of neighboring values K from 1 to 10, and the PM method includes the annual and semi-annual Periodic information of ZTD. The results show that for K=1, the missing rate is 10%, 20%, and 30%, and the mean RMS values of KNN/PM-ZTD and GNSS ZTD are 9.9/7.2, 14.3/10.3, 16.7/12.1mm, respectively, and the RMS difference between KNN/PM-ZTD and GNSS ZTD increases with the value of K from 2.7 to 0.5/4 to 0.7/ 4.6 to 0.7 mm. Therefore, it is verified that the PM method is superior to KNN, and the present method is important for the study of the complementary long-time series data.