This PDF file contains the front matter associated with SPIE Proceedings Volume 12129 including the Title Page, Copyright information, and Table of Contents.

Global land cover product and extraction technology with remote sensing images are the common methods to obtain the information of surface water. This paper selects the GLC_FCS30-2020 (GLC-2020) product and two results of water extraction by Sentinel-1A SAR images on different date and analyzes the consistence and accuracies of them. Verifying by the experiments, the consistence between the results of GLC-2020 and SAR method is higher than 80%, and the values of UA and PA of them are all more than 90% and 85%, respectively. The three results of surface water extraction all meet the needs of general applications. Considering the convenience and continuity, GLC-2020 is adopted in more applications and the SAR method is more suitable for time series observation.

Remote sensing image scene classification (RSSC) is a hot topic in remote sensing. Aiming at the problem of remote sensing image of limited labeling samples and class-imbalanced, we proposed a RSSC framework based on efficientnetB7. The framework uses mirroring, rotation, cropping, HSV disturbance, and gamma transform to improve the problem of class-imbalance, and restricts the rotation angle of the high-rise-sparse-buildings to make it in line with the actual situation., and then the pre-trained model is used for training. The results show that the kappa and OA of the model increased 0.139 and 0.055, respectively, and the classification deviation caused by class-imbalance is alleviated.

Ship detection in remote sensing images is a challenging and important task. Many methods of deep learning have been widely used in this field in recent years and most of them rely on a lot of predefined rotated anchor boxes, which not only lead to inaccurate angle predictions but also introduce excessive hyper parameters and high computational cost. In this paper, we propose an anchor-free rotation ship detector (ARS-Det) using axis-based representation scheme to address these issues. Our ARS-Det regards the detection of arbitrary-oriented objects as predicting the points at both ends of the longest axis and width of the objects. Attention FPN model is designed to capture multi-scale and key features of arbitrary-oriented target. Then, points assignment and orientation center-ness computation are proposed to screen of positive samples accurately. Experiments on public ship dataset HRSC2016 show that our method achieves state-of-art performance on ship detection in remote sensing images.

The black-odor water has significant negative impacts on the sustainable development of society and survival of mankind. Taking the main rivers in the urban built-up area of Fuzhou city as the research object, we obtained 15 normal water samples and 45 black-odor water samples on the remote sensing image of Gaofen-2(GF-2) on December 7, 2016 according to the list of black-odor water bodies announced by the municipal government. The paper puts forward a new Black and Odorous Water Index (BOWI) model that based on the ratio of band 2 and 3 to band 3 and 4 to identify the urban blackodor water body based on the reflectance spectrum on image. And the model is validated by the observation and the producer’s accuracy can be up to 71.43%. Result shows that black-odor water sections are widely distributed but discontinuous, and they are concentrated in the densely populated areas of urban areas. Domestic sewage, industrial waste water and broken waterfront are the main reasons.

Some ancient murals were found to be repainted on the surface of the original murals to form multi-layer murals. The patterns on the original layer are of great significance for studying the social and cultural behavior in that time. The hyperspectral imaging covers the visible and near-infrared bands, which has advantages for the information extraction of multi-layer murals. Therefore, a method to study the transmission performance of hyperspectral imaging on multi-layer simulated mural samples is proposed. By making mural samples, the mineral pigment painted on the surface is covered with 0-11 different layers of lime water. Then the samples were collected with hyperspectral images, and the method of principal component analysis and band calculation were used to analyze the enhancement effect of the mural patterns covered by different layers of lime water. The results show that hyperspectral imaging has certain transmittance to the interior of the mural and can enhance internal pigment information. The research results can support the information extraction of multi-layer murals to some extent.

Remote sensing can be used to monitor snow cover in a large area, which has a major contribution to climate and environmental research, the establishment of hydrological models, and disaster prevention and mitigation. Multiple remote sensing data fusion is an important means to improve the efficiency of snow cover detection. This study used 192 SNOTEL stations in the northwestern United States to evaluate the commonly used MODIS datasets and IMS snow and ice product. It is found that the snow inversion effect in the cloud-free area of the MODIS datasets is better, and IMS snow and ice product has lower snow recognition rate but no cloud. Based on this, we integrated the snowy area of the MODIS datasets into the IMS snow and ice product, and generated a new product called MIMS. The results show that the SRP of MIMS is 72.93%, which is higher than the MODIS datasets (72.78%) and the IMS snow and ice procduct (70.46%). At the same time, the new product eliminates cloud pollution and improves the spatial resolution. It can be seen that the fusion of multiple remote sensing data can effectively improve the performance of snow inversion, and has important reference value for the accurate monitoring of snow in different regions.

At present, there are many methods to establish 3D models, but mainly through traditional measurement methods, so that there are difficulties in point cloud modeling of indoor spaces of buildings and underground buildings. Nowadays, indoor 3D modeling is gradually becoming the focus of people, this paper takes an underground parking garage as a practical case, uses SLAM technology to scan to get point cloud data, and combines Geomagic Studio, AutoCAD, SketchUp and other software to effectively process the data and reconstruct the 3D model of the building with reference to it.

With the advancement of science and technology, image processing technology is widely used. In this paper, we use computer technology to process the image data of the screenshots in the classroom monitoring to achieve the goal of extracting human heads. In this study, we selected screenshots of different size classrooms, camera views, and classroom lighting conditions. Combined with the PASCAL VOC dataset format, we applied the Labellmg-master image annotation tool to annotate the human head. The traditional YOLOv3 algorithm misses and repeats detections. Its accuracy and recall rates are poor. Therefore, we propose an optimization algorithm for classroom monitoring via head detection based on the YOLOv3 algorithm. The main improvements are the clustering the dataset labels with the Kmeans algorithm, increasing the size of the output feature map, and optimizing the loss function. Experiments were conducted with both the YOLOv3 and optimized YOLOv3 algorithms. By comparing the experimental results, we can conclude that the optimized YOLOv3 algorithm can improve the detection accuracy by 6% and recall rate by 11% on the test set, which has a high detection performance. This shows that image data processing technology has a better application effect in classroom monitoring.

Surveillance videos have been deployed throughout the "Safe City" construction to employ large-scale surveillance video information to cope with problems in security and other sectors. So how to study into foreground target extraction and image processing in surveillance video, might give useful help to generic video processing problems. Aiming at extracting foreground targets from surveillance videos with various backgrounds and targets, a matching foreground target extraction model was developed. First, a static background extraction model based on the enhanced three-frame difference technique was created using the adaptive iterative threshold method. Second, by combining the standard twoframe difference approach with the Gaussian mixture model, a dynamic background extraction model based on the neighborhood search method was developed. Finally, a dynamic backdrop extraction model for unstable shooting was created using the enhanced three-frame difference approach, corner feature matching, and linear affine transformation. The experimental results demonstrated that the model can efficiently recognize and extract the foreground targets of surveillance video, which solves the problem of extracting complicated targets.

Synthetic aperture radar (SAR) image registration using the histogram of orientated phase congruency (HOPC) is a challenging task due to HOPC lacks rotation and scale invariance. In this paper, a modified HOPC algorithm is proposed to address the aforementioned problems. Firstly, the brief review on HOPC and its limitations are introduced. Secondly, the scale invariant feature transform (SIFT) algorithm is used to rotate image for handle the rotation invariance. Subsequently, multi-scale phase congruency features are extracted in phase congruency scale space, and original HOPC is used to match features. Finally, we use random sample consensus (RANSAC) to filter all the matches to obtain correct matches. In this paper, TerraSAR-X and Sentinel-1A data are used as experimental data. We utilize different assessment paraments to evaluate the performance of the modified HOPC algorithm. The experimental results demonstrate the proposed algorithm outperforms the state-of-art algorithms in terms of registration accuracy.

The automatic extraction of the roads from the remote sensing images is significant for the monitoring of the sandstorm hazards to the traffic arteries of the desert area. Since the current pixel-based road or the object-oriented extraction methods are easy to cause the noises or the adhesion phenomena, the U-Net deep learning network is applied to extract the desert roads from the Google Earth, GF-2 and JiLin-1 image datasets in this paper. Firstly, in order to improve the generalization ability of the U-Net network model, the datasets are expanded by means of rotating, mirroring, contrast stretching, and intensity dithering. Then under the constraints of the hyperparameters, the U-Net model is built and trained until the loss function value tends to be stable. Finally, the U-Net algorithm is adopted to extract the highways pass through the Takramakan desert and the Kumtg desert areas in western China. The experimental results demonstrate that the U-Net algorithm is efficient and performable.

This study investigated the dynamic changes of land use/cover in Guangdong-Hong Kong-Macao Greater Bay Area (GBA) from 1979 to 2016 by using transfer matrix method combined with remote sensing and GIS. The results show that during this period, Farmland reduced 19744.27 km² with a drop of 81%, 6147.76 km² of which had been designated for industrial land. The land use changed variously among the districts in the GBA, and the comprehensive dynamic index of land use of Zhongshan city was the highest; and that of Huizhou city was the lowest. Social and economic factors such as population, economy, transportation and policy all have an important impact on the land use pattern and its changes in the GBA.

Precipitation is an important factor that predicts the occurrence of forest fires in the future. This study uses a time decay model to calculate the comprehensive precipitation index, which is an exponential weight decay model. This method can better represent the effect of precipitation in predicting the occurrence of forest fires. Besides, this study used the Support Vector Machine (SVM) regression model to construct a forest fire warning model. In the same area, using the comprehensive precipitation index compared with the average precipitation, the accuracy of the three forest areas in the test set has been improved by approximately 5%.

The evaluation of forest fire damage degree is a particularly important operational need in carrying out forest fire rehabilitation work. Comparing to traditional evaluation methods, satellite remote sensing is faster and more effective with a larger field of view, which is supposed to be used to evaluate the degree of forest fire. The paper takes the forest fire of Datong Volcano Group Geopark on April 30, 2020 as the research object. Firstly, sentinel-2 satellite images were acquired before, during and after the fire periods, and the images were band synthesized and calculate the remote sensing index, and finally the change detection method was used to analyze the extent of forest damage after the fire. The results show that the combined short-wave infrared and near-infrared bands of sentinel-2 satellite can be used to detect the origin of fire and fire line. Normalized Burn Ratio (NBR) shows fire area intuitively and Differenced Normalized Burn Ratio (dNBR) shows how severe the forest is damaged. After the fire, the slightly damaged area is the largest and the extremely severely damaged area is the smallest and the damage evaluation can provide an effective technical method for subsequent forestry vegetation restoration and forest fire prevention.

With the completion of the China BeiDou-3 satellite constellation, the available signal sources for GNSS-R ocean altimetry have been further enriched. At the same time, the BeiDou-3 satellites have adopted more superior signal structures, which provide the possibility to improve the GNSS-R ocean altimetric precision. To analyze the influence of the different signal structures of BeiDou-3 satellites on the GNSS-R ocean altimetric precision, this paper has carried out the following research. Firstly, the modulation modes and autocorrelation functions of the BeiDou-3 satellite signal are analyzed, and the principle of ocean altimetry is presented. Secondly, the reflected delay power waveforms corresponding to different signal structures are simulated. Thirdly, the ocean altimetric precision based on different signal structures is compared and analyzed by using simulated reflected waveforms. The results show that compared with the traditional GPS L1 band C/A code altimetry, the ocean altimetry based on the BeiDou-3 satellite signal will obtain higher precision.

The dynamic detection of coastline is important for the study of sea level rising caused by global climate change. Therefore, it is especially meaningful to select an appropriate method to extract the coastline accurately. In this paper, the coastal zones of Reykjavik, Iceland was chosen for the study area, and the method of combination water indices (i.e., NDWI, MNDWI, EWI, and AWEI) and the maximum interclass variance method (Otsu) was used to automatically detect the coastline. The results show that the NDWI and EWI can accurately extract the coastlines in this region, and NDWI performs better than others (accuracy of 91%). The conclusions obtained in this paper have the potential to be applied to coastline extraction in the polar regions.

The investigation of emergency environment of flood and waterlogging disaster focuses on real-time and timeliness, which is the need of disaster analysis and emergency rescue.The airborne three-dimensional laser scanning system (LiDAR) is flexible and less affected by the weather. Moreover , it has the characteristics of being able to penetrate the leaf canopy of vegetation directly to the ground .Based on these characteristics, liDAR technology is proposed to obtain detailed topographic and geomorphic data in small area.Through the investigation and study of small watershed in the south of Jinan, this paper systematically introduces the working principle of airborne LiDAR, data processing process, and the specific process of obtaining high-precision DEM and surrounding topography.Application cases show that by means of airborne lidar to obtain small watershed basic geographic information, the data processing speed and the precision and quality of data and subsequent digital products can meet the needs of flood disaster emergency investigation.It provides basic geographic information data for analyzing flood and mountain flood prevention and control work caused by short-term flood.

In order to comprehensively understand the main reasons for the damage of Shuangjiantu seawall under the influence of strong typhoon, based on the analysis of a large amount of marine hydrographic data monitored during the active period of Typhoon No.6 “In-fa” in 2021, the marine environment around Shuangjiantu seawall in Zhoushan was projected and the basic situation of Shuangjiantu seawall design and construction was analyzed to find by analyzing the basic design and construction of Shuangjiantu seawall, we found out the damage characteristics and mechanism of typhoon “In-fa” on Shuangjiantu seawall in Zhoushan. It was found that the seawall was most affected when the seawall orientation overlapped with the strong wind direction. The reason for the damage of Shuangjiantu seawall: the typhoon hit the seawall head-on after landing, and the infinite wind area in front of the seawall constituted a bad marine environment condition in front of the seawall. Although the overall design of the seawall met the standard, but did not consider the impact of local water depth topographic differences on the seawall, different sections at the same structure and strength. The water depth at the larger wave is not easy to break, there is a greater force on the seawall, so the seawall damage mainly appeared in the dike in front of the local water depth deeper location.

This paper first discusses the concepts and techniques of data mining. Then the sources and effects of indoor environmental pollution are introduced, from which the data to be monitored in indoor environments are derived and classified, and four approaches to data acquisition are discussed. Five main application directions of indoor environmental monitoring research are then discussed, including: environmental data visualization, indoor safety warning, building energy consumption, indoor environment and individual health analysis, and research on indoor environmental management decisions. Finally, we compare the research content, potential problems, and future research hotspots in this field.

In order to achieve intelligent classification of air pollution-related microblog texts with semantically mixed public perception, with the help of the Word2Vec model developed by Google, this paper uses a large number of microblog texts as the training set for word vectors training, the trained word vectors are embedded in convolutional neural network (referred to as CNN) model. Based on the established classification criteria of gaseous pollution, dust pollution, smoke pollution and haze pollution, the 1000 air pollution-related microblogs were screened out to design and implement an automatic and intelligent classification algorithm. The experimental results show that the overall accuracy rate is 94.75% and the overall recall rate is 94.50%, which can meet the accurate classification of air pollution-related microblog texts, and provide accurate corpus data for later analysis of the semantic features of air pollution topics and the public's perceived emotional intensity.

High spectral analysis technology has the advantages of fast, accurate and nondestructive, and is widely used in the field of leaf nitrogen analysis. In order to explore the optimal inversion model for monitoring the nitrogen content of Phyllostachys pubescens. The collected Phyllostachys pubescens sample data were divided into modeling set and validation set based on SPXY (Sample Set Partitioning based on Joint X-Y Distance Sampling) method and Random method, respectively. The SPA (Successive Projections Algorithm) was used to extract the characteristic wavelengths of the original and transform spectra. And the vegetation index and red edge parameters with high correlation with the nitrogen content of Phyllostachys edulis were selected. Then the PLSR estimation model based on the nitrogen content of Phyllostachys edulis was established. The results showed that compared with the random sample partition method, the SPXY sample partition method increased the estimation accuracy R² by 0.13 on average, reduced RMSE by 0.50 on average, and increased RPD by 0.58. The PLSR estimation model of CR-FDR established had the highest fitting accuracy of N content in Phyllostachys pubescens, R² was 0.85, RMSE was 1.32, RPD was 2.42. The inversion model combined with UAV hyperspectral monitoring data can better reflect the spatial difference of Phyllostachys pubescens nitrogen content.

Since 2000, China has witnessed rapid social and economic development and rapid improvement of urbanization level. Analyzing the driving force of urban construction land expansion is an important basis for predicting the future expansion of urban construction land. This study uses EVI, DMSP/OLS and Luojia1-01 nighttime remote sensing data products to extract the urban construction land range of Jinan from 2000 to 2018. The evolution process of urbanization is characterized by three indicators: AGR, elastic coefficient and city compactness. The migration of urban center is analyzed by standard deviation ellipse. Combined with the statistical yearbook data over the years, the driving mechanism of urbanization is analyzed by using geographic detector model from the perspectives of economic construction, social development, infrastructure and environment. The results show that: since 2003, Jinan city has a trend of scattered expansion, and the overall expansion direction is obvious. From 2003 to 2008, the urban expansion speed is faster, followed by 2013 to 2018, and the expansion speed is slower from 2008 to 2013. The urban compactness decreases with the increase of construction land area, and the urban center move to the east by slightly north. The urban expansion is mainly driven by economy, mainly in GDP growth, industrial structure optimization, economic opening policy, followed by population and transportation facilities, and there is a two-factor enhanced relationship between the indicators.

Murals are an important part of China's cultural heritage that have high historical, scientific and cultural values. The traditional methods of extracting the diseases of murals are mainly artificial measurement and orthographic drawing, which are inefficient for the rapid statistics of large-scale mural diseases. To solve the above problems, a disease data set was established based on mural orthophoto images. And the image deep learning YOLOv4 algorithm was used to train the data set. Through comparative experiments, the most suitable method for YOLOv4 network detection was found to make the data set, so as to realize automatic rapid recognition of mural orthophoto images and express the disease information. Through experiments, it is proved that the accuracy of disease identification by this research method reaches 86.51%, and the extraction results can provide favorable data support for scientific and technological protection of cultural relics.

Knowledge graph is a structured organization form of data, and it will be useful to introduce it into the tourism recommendation system as a kind of auxiliary information. The research proposes a model of attractions recommendation based on knowledge graph, which aims to use the semantic information and network structure of knowledge graph to encode the potential interests of tourists. Specifically, the model uses graph convolutional neural network to spread the embedding representation of attractions, and considers the importance of relationship and entity similarity in the convolution process to reflect the difference in preference of tourists. In addition, we also use the attention network to encode the sequential movement of tourists. The study developed a Beijing tourism knowledge graph to organize and share travel information, and used travel notes data to verify the model’s performance. Experimental results show that the recommendation model based on tourism knowledge graph can effectively overcome the problem of data sparsity and achieve better performance than state-of-the-art models.

In recent years, ecological security issues have been complex and diverse, and grassland ecology has also faced problems such as overgrazing, land desertification, and grassland degradation. The protection of grassland ecology is an important issue. As an important biological resource of grassland ecology, grassland plants play an important role in preventing wind and sand fixation and conserving water and soil. Most of the current related researches are concentrated on the ecological function of certain plants or the category of botany. Therefore, this paper analyzes the composition and relationship of grassland plants on the basis of analysis of previous studies, and designs a conceptual framework for grassland plant knowledge graph construction using an ontology-based method; Using a template-based method and DeepDive knowledge extraction tool to extract knowledge from a large amount of grassland plant data, construct a grassland plant knowledge graph. The knowledge graph constructed in this paper covers the conceptual attributes, inter-species relationships and plant functions of grassland plant entities from a macro perspective. It can help protect grassland plant resources and provide basis and services for grassland ecosystem monitoring, protection and management.

The RBN-DGNSS, as a safeguard facility to support the navigation of ships, has been shut down by some countries, while the stations in China have been upgraded and can integrate and broadcast Beidou differential signals. This article analyzes the specific reasons and countermeasures for shutting down radio beacon differential stations in a few countries, as well as the guidance opinions of IMO and IALA on the development of RBN-DGNSS. Combined with the new characteristics of China's RBN-DGNSS differential station after upgrading and transformation, this paper evaluates the risks of shutting down the existing RBN differential station to China's maritime location service, user usage, international compliance and other aspects, and draws a conclusion that it is not appropriate for China to shut down the differential station completely at the present stage. Suggestions on future development trends of China's RBN-DGNSS combined with Beidou Satellite Based Augmentation System(BDSBAS) and other technological development trends, to provide reference for the development and transformation of RBN-DGNSS in relevant countries in the world.

With the transformation of urban functions and the development of tourism, the research on the temporal and spatial behavior characteristics of tourist flow and destination preference within cities has gradually attracted the attention of domestic and foreign scholars. Based on the digital footprint data, with the help of social network analysis and mathematical statistical analysis, the time and space distribution and network structure characteristics of the tourist flow of external tourists in Hangzhou are studied. The results show that: Hangzhou’s tourist flows have strong seasonal time characteristics; it presents a “dual core” and gradually decreasing spatial distribution characteristics, as well as the spatial preference characteristics dominated by natural scenery and historical sites; the overall density of the tourist flow network is relatively loose.

Water conservation is one of the ecological service functions of terrestrial ecosystem, which is of great significance for monitoring and protecting the ecological environment. The area with Danjiangkou reservoir as the core is an important ecological service function area of water conservation in China. Aimed at the distribution and dynamic change of water conservation function in this region, this paper inverses the regional water conservation capacity based on the water balance principle in 2015 and 2020, also studies the spatial difference and interannual variation in combination with the changes of land use, precipitation and temperature. The results show that the type of land use plays a decisive role for the maintaining and improving of regional water conservation, and the climatic conditions such as precipitation and temperature have a great impact on the fluctuation of water conservation capacity.

The purpose of pansharpening is to fuse low-resolution multispectral image (LRMS) and high-resolution panchromatic image (PAN) to obtain high-resolution multispectral (HRMS). In response to the shortcomings of traditional remote sensing image fusion algorithms causing spectral distortion, more and more deep learning algorithms are utilized, and this paper proposes a new deep network structure, two-branch Self-Attentive DenseNet network. In terms of maintaining high spatial resolution, the image feature information is extracted by different inch-scale convolutional kernels, and the effective feature information is enhanced to suppress the invalid image information by using DenseNet network model and introducing Self-Attention, and the fused image spectral information is enhanced by using hopping connection to maintaining the spectral structure. Experiments show that the proposed method of this paper has improved image quality evaluation metrics compared with the previously existing traditional fusion algorithms and deep network algorithms.

The Internet and social media have become important carriers of destination image dissemination, and the photos on social platforms reflect to a certain extent tourists’ perception preferences in tourist destinations. This paper uses the photos actively uploaded by Xi’an tourists on the Liangbulu website as the research data source, and uses the convolutional neural network model to identify the photo scene, then analyzes the tourism image of Xi’an and its spatial distribution characteristics from the perspective of tourists’ perception. The study shows that the spatial distribution of Xi’an’s building, urban landscape, place/region, transportation, interpretation, night scenery, and water tourism image is generally “concentrated in the city center, sparse and isolated in the periphery”; the people and nature tourism images generally show the characteristics of “sparse and scattered” spatial distribution.

Aiming at the particularity of China's energy security in the context of "carbon neutrality" and "carbon peaking" with reference to EISD, this research constructs China's energy security indicator system covering three dimensions of society, economy and environment. The assessment relies on principal component analysis to evaluate China's energy security indicator from 1985 to 2017 and uses the ARMA model to predict the development trend of China's energy security indicator from 2018 to 2025 at the same time. The empirical results show that China's energy security indicator expressed a U-shaped development from 1985 to 2025 and China's energy security indicator showed a positive growth trend from 2018 to 2025. The conclusions and recommendations are as follows: Although China's energy security indicator is showing a trend of positive growth, there is still a large room for improvement in the overall level of China's energy security; the general industrial structure needs to be optimized and adjusted; and the implementation of environmental protection work should be further strengthened. This research has a positive contribution to environmental monitoring as well.

In view of the city small and medium-sized bridge potential damage area detection efficiency is not high shortcomings, using the ground three-dimensional Laser Scanner (TLS) point cloud data, determine the potential damage area of small and medium-sized span bridge, and accurately detect the key damage location of small and medium-sized span bridge. In this paper, a set of accurate potential damage detection process is proposed, and the process is applied to the Beisha Bridge in Beijing city for potential damage detection analysis. It is divided into three steps :(1) quantification of bridge deck damage by applying the range-gradient algorithm model; (2) feasibility of detecting potential damage of bridge deck by gaussian curvature of point cloud data is analyzed in detail, and the model is constructed; (3) abnormal areas of bridge deck change are identified by taking the experimental data of beizan bridge as an example. The experimental results show that the proposed Gaussian curvature damage identification method is consistent with the displacements and gradients method, which can be applied to the analysis of potential damage mechanism of Bridges.

Port competitiveness refers to the port's ability to compete for various resources, which reflects the port's position in the region. A correct assessment of port competitiveness will help to better promote the coordinated development of ports and enable ports to participate more deeply in the national development strategy plan, thereby contributing to the sustainable development of ports and hinterland cities. Based on the Automatic Identification System (AIS) data, this paper uses the complex network method to calculate the complex network indicators of 11 urban ports in the Guangdong-Hong Kong-Macao Greater Bay Area, and then uses Borda Count to rank the port competitiveness of the 11 ports. The results of the study show that the ports of Hong Kong, Macao, Guangzhou and Shenzhen in the Greater Bay Area have superior positions and are highly competitive in the internal and external conditions of the ports and in the route network. They are important hubs for the “Belt and Road” construction. Other countries or the port can give priority to cooperating with it. The ports of Jiangmen, Zhongshan and Zhaoqing, which are less competitive, can enhance their competitiveness by improving the investment environment and port operation capabilities.

Fast and accurate sea ice detection is of great significance to the natural environmental protection of the sea and the development of the marine economy. With the development of satellite remote sensing technology, sea ice detection based on high-resolution SAR images has received wide attention. Given the serious pretzel phenomenon of image-level classification results and the limitation of object-level classification by segmentation scale, this study proposes a sea ice extraction method based on the coherence and magnitude information of TanDEM-X images and combining image level and object level. The method was compared with the sea ice extraction results of the traditional method, and the results showed that the overall accuracy, user accuracy, product accuracy, and Kappa coefficient of this newly constructed sea ice extraction method improved 38.74%, 19.43%, 37.48% and 0.7597, respectively, compared with the traditional extraction method, which significantly improved the sea ice extraction accuracy.

In this paper, the atmospheric correction is performed using the three methods of 6S, FLAASH, QUAC and compared with the MOD09A1 product and the measured spectral data. The experimental results show that the three methods have maintained better consistency with the measured spectral data, which can effectively eliminate the influence of the atmosphere in radiation transmission, effectively restore the real surface reflectivity, the correction effect is 6S< FLAASH< QUAC. The experimental results can provide a reference for atmospheric correction processing of GF-1 WFV multi-spectral images.

This Flood disasters, with their high frequency and major hazards, seriously endanger the safety of human life and property. Therefore, flood monitoring is of great importance. Synthetic aperture radar (SAR) is not affected by clouds and rain, and obtain effective data for flood monitoring. In July 2021, Weihui City encountered heavy rainfall, causing severe flooding. This paper selects the Sentinel-1A SAR data of Weihui City before the flood (July 15), during the flood (July 27) and after the flood (August 8), and uses the object-oriented threshold method to extract the water body information, and conduct the flood inundation area monitoring and analysis. The results demonstrate that the use of Sentinel-1A data based on the object-oriented threshold method can achieve rapid monitoring of flood areas. Before the flood occurred in the main urban area of Weihui City, the water body coverage area is 4.18 km² , and the water body coverage area is 45.72 km² during the flood disaster. After the flood receded, the coverage area of the water body is reduced to 15.98 km² . This indicates that the method proposed in this paper can effectively extract the coverage of water body and provide a reliable technical means for flood monitoring.

The development of Internet information technology has promoted the emergence of many social media and interactive platforms, which contain a large amount of POI information related to tourists' travel, providing a new source of big data and perspective for studying the spatial distribution characteristics and spatio-temporal change characteristics of tourists' POI. In this paper, we use Flickr image sharing website to obtain POI data of inbound tourists in Tibet, construct a tourism flow network structure evaluation system from the perspective of social networks, explore the characteristics of spatial and temporal distribution of POI of inbound tourists in Tibet, and then make the following suggestions: 1. The structure of inbound tourism flow in Tibet is loose, and the connection between attractions is weak. With the development of transportation infrastructure, the tourism attractiveness of northern and eastern Tibet is gradually rising.2, cold attractions have unique attraction to inbound tourists, and creating special towns can help promote inbound tourism development.