This study constructs a water quality inversion model based on mathematical statistics and machine learning methods by combining water quality parameters, spectral data collected on June 15, 2023, and synchronous Sentinel-2 imagery data to quantitatively invert the Chl-a concentration in Luoma Lake. Through comparative analysis, the model with the best performance is selected to analyze the Chl-a status in Luoma Lake. The study found that Chl-a shows a high correlation with the B5 and B9 bands of the Sentinel-2 imagery, and this correlation is further enhanced after band combination processing. In the Chl-a inversion model, the FA-SVR model demonstrates the highest accuracy (R²=0.86, RMSE=2.77, MAE=2.10) compared to traditional mathematical statistical regression models and other machine learning models (FA-RF, FA-XGBoost). The inversion results reveale that the nearshore area in the northeastern part of Luoma Lake has higher Chl-a concentrations, which may be related to the presence of fishpond farming and high eutrophication in the northern waters. This study highlights the significant application value of machine learning technology in improving the accuracy of water quality remote sensing inversion, provides important technical support for water quality monitoring and management in Luoma Lake.

This article inverts the water quality parameters of the key water bodies called “one river and three lakes” in the demonstration zone of integrated development of the Yangtze River delta using airborne hyperspectral imagery. The experiment collected the full-spectrum hyperspectral image data in the demonstration area, and obtained the water-leaving reflectance of the water bodies after image preprocessing. The water samples and water spectra are collected synchronously at the flight time. By using the competitive adaptive reweighted feature selection method, the optimal bands are selected for establishing the inversion model, and the water quality thematic mapping is also completed. The coefficient determination R² of the validation set is more than 0.6, and the data error with the water quality monitoring station is less than 10%. This study achieves precise monitoring of water quality parameters in the demonstration area, confirms the application value of airborne hyperspectral remote sensing in the inversion of water quality parameters. The results could enable high quality for the development of the demonstration area.

As one of the top ten invasive plants in China, water hyacinth has a strong reproduction ability, and it has flooded in plain river network, seriously affecting the aquatic ecology and river flooding safety. In this paper, taking Xinghua city as an example, based on Sentinel-2 images from 2017 to 2021, the distribution of water hyacinth in the river network is automatically extracted by the support vector machine algorithm, the average coverage of water hyacinth in 59 major rivers in Xinghua city is computed, a hierarchical classification method for water hyacinth outbreaks is established, and the spatio-temporal heterogeneity of NDVI values in the water hyacinth coverage area of Xinghua city is detected. The experimental results show that between 2017 and 2021, the percentage of water hyacinth area ranged from 0.05 to 0.2, with the lowest value of 0.084 and the highest value of 0.176; there are 9 rivers with severe outbreaks, 34 rivers with moderate outbreaks, and 16 rivers with mild outbreaks according to the outbreak class; and the spatio-temporal heterogeneity detection indexes showed a general upward trend, and reached a peak value of 1.025 in the second half of 2020. Detecting the spatio-temporal changes of water hyacinth through spatio-temporal heterogeneity can provide a scientific basis for monitoring and rational control of water hyacinth outbreaks.

Based on the statistical, territorial survey and image data of 2009,2012,2015,2018 and 2021, the entropy weight method and TOPSIS model are used to collaborative analysis the water resources carrying capacity of 57 county-level administrative regions in the Yellow River basin(Gansu section). The results show that: ①From 2009 to 2021, the number of county-level administrative districts with low levels of water resources subsystem, socio-economic subsystem and ecological environment subsystem in the Yellow River basin (Gansu section) decreased significantly, while the number of high levels increased. ②The regions of the Yellow River basin (Gansu section) with unsatisfactory water resources carrying capacity are concentrated in Longzhong and Longdong, while the regions of Maqu, Luqu, Zhuoni and Xiahe county with medium ideal, relatively ideal and ideal water resources carrying capacity are concentrated in Gannan plateau. ③During the study period, the water resources carrying capacity in most regions of the Yellow River basin (Gansu section) is stable or increased. This also reflects the remarkable results of environmental governance in the past two decades, and the steady improvement of water conservation capacity and comprehensive utilization efficiency of water resources.

The rapid accumulation and growth of water quality monitoring data brings new opportunities and challenges to water quality research. In view of the problems of poor timeliness, unintuitive data display, and inefficient data processing faced by water quality monitoring and analysis at present, this paper constructs WaterSpark, a distributed processing framework for real-time water environment monitoring and big data analysis, based on big data technology and data visualization technology, using the improved Canadian Council of Ministers of the Environment water quality index (CCME-WQI) and the Spark machine learning library (MLlib), Applying the water quality monitoring data of nine plateau lakes in Yunnan province, the results show that WaterSpark has excellent performance in real-time water quality transmission, cleaning and archiving, and efficient computing and analysis. It can enable large-scale water quality data to be captured and analyzed timely and accurately, and the distributed data sets and clusters can cope with the growing water quality data to ensure performance scalability, and to support more water quality indicators and water quality monitoring on a larger scale.

Forest above ground biomass is an important index to measure the carbon sequestration capacity and productivity of forest ecosystem. Rapid, continuous and large-scale inversion and mapping of forest above ground biomass could be accomplished by remote sensing. How to establish an extrapolation model to estimate forest biomass more accurately is the key to remote sensing inversion of forest biomass. Taking Chongyang county, Hubei province as the research object, the above ground biomass mapping of forest growing season from 2013 to 2023 with 30m resolution time series was completed by using spaceborne LiDAR ICESat-2 data and different types of forest biomass modeling, combined with multi-spectral Landsat 8 satellite images, and using multiple linear regression and random forest algorithm. The results show that ICESat-2 ATL08 data could estimate the spot biomass well, and the R² of coniferous forest is 0.872, and the R² of broadleaf forest is 0.806. The experiment verifies the ability of space-borne LiDAR to estimate the biomass of different forest types, and provides a method and basis for the future study of the spatio-temporal change pattern of forest biomass and carbon sink at large regional scales.

In order to address the limitations of using global navigation satellite system (GNSS) or Gravity Recovery and Climate Experiment (GRACE)/GRACE follow-on(GFO) technology alone for retrieving regional terrestrial water storage changes. A method for integrating GNSS and GRACE/GFO data to estimate changes in regional terrestrial water storage is provided in this paper, drawing inspiration from the “decomposition-reconstruction method.” Changes in terrestrial water storage in southwestern China from 2012 to 2022 are estimated by integrating GNSS and GRACE/GFO data using time series data from 90 GNSS stations in southwestern China. The performance of our fusion method is verified through spatial distribution analysis and comparison with other fusion methods and precipitation data. The results demonstrate that the fusion inversion combines the advantages of GNSS and GRACE/GFO in spatial distribution. On the time scale, the inversion results from GNSS and GRACE/GFO both lag behind precipitation by 2 months, whereas the fusion inversion results lag behind precipitation by only 1 month, and the corresponding correlation coefficients are also higher than those of GNSS or GRACE/GFO inversion results. Therefore, the fusion method provided in this paper offers a reference for combining multiple data sources to obtain more reliable regional terrestrial water storage changes.

Henan province is one of the regions in China with relatively severe water scarcity. The excessive consumption of groundwater has become an important factor restricting the economic and social development and ecological environment protection in the region. Based on GRACE, GRACE-FO satellite gravity data, and GLDAS land surface assimilation data from 2003 to 2022, this study analyzes the temporal and spatial changes of groundwater reserves in Henan province over the past 20 years using singular spectrum interpolation and time series decomposition methods, as well as the influencing factors. The results show that the variation of groundwater reserves in Henan province over the past 20 years shows a V-shaped trend, with the rate of change decreasing stepwise from the southwest to the northeast. The rate of change before and after 2015 is -15.53 and 8.88mm/a, respectively, and the groundwater in the northeastern region has shifted from deficit to surplus. A comparison of the inversion results with precipitation and water transfer data from the south-to-north water transfer project reveals that precipitation is closely related to the change in groundwater, but it is not the main reason for the long-term decline in groundwater. The replacement of groundwater pumping by deploying surface water for production and living in the receiving area through the south-to-north water transfer project effectively alleviates the groundwater deficit, playing a crucial role in groundwater recovery. The study results provide an important reference for understanding the spatial and temporal changes of groundwater reserves in Henan province over the past 20 years and for better leveraging the role of the south-to-north water transfer project in water management.

Desertification is one of the important ecological and environmental problems that threatens regional development and human survival. Therefore, desertification monitoring is an important basic work in desertification prevention and control. This article takes the Datong coalfield in Shanxi as an example, uses the NDVI-Albedo feature space to construct the desertification difference index (DDI) from 2000 to 2021, monitors its spatiotemporal changes in desertification, and finally uses geographical detectors to analyze the influencing factors of desertification. The results show that from 2000 to 2021, the desertification of Datong coalfield changed from severe to mild, the area of severe and extremely severe desertification decreased significantly, and the area of mild desertification increased significantly. Overall, 57.75% of the regions showed a significant improvement in desertification, but at the same time, 0.03% of the regions showed an aggravating trend. The results of geographical detection show that vegetation is the main factor affecting desertification, and the impact of vegetation on natural factors such as rainfall and elevation has increased significantly. At the same time, the impact of vegetation on human activities such as land use and population density cannot be ignored.

Glaciers and glacial lakes are well developed in the Koshi River basin. With global warming, glacial meltwater has become an important source of glacial lake expansion. At the same time, it will cause natural disasters such as glacial lake outburst flood (GLOF). Therefore, it is significant to study the temporal and spatial changes of glacial lakes in the region. In this paper, based on Landsat series remote sensing images and digital elevation model, the temporal and spatial distribution and variation characteristics of glacial lake area in the Koshi River basin From 1990 to 2020 are analyzed. The results show that: ① From 1990 to 2020, the glacier area in the Koshi River basin showed an accelerating trend of retreat and the trend of accelerated retreat has been further strengthened in the past ten years. ②The glaciers in the Koshi River basin are mainly distributed at an altitude of 4800～6800m, with a slope of 5～20°.③ From 1990 to 2020, the glacial lakes in the Koshi River basin have shown an accelerated expansion trend. The area of glacial lakes has expanded by 26.09km². The glacial lakes with an area of >0.25km² expanded rapidly. The area of glacial lakes connected to glaciers has expanded significantly, with an area growth rate of 49.39%.④ From 1990 to 2020, a total of 11 glacial lakes in the Koshi River basin have burst, and some glacial lakes have completely burst or repeatedly burst. The area of glacial lakes is 0.02～0.7km². ⑤From 1990 to 2020, the temperature and precipitation in the Koshi River basin showed a fluctuating growth trend resulting in accelerated retreat of glaciers, accelerated expansion of glacial lakes, and frequent occurrence of glacial lake outburst flood in the Koshi River basin.

In recent years,several new subway lines have been added in Nanjing,in order to explore the hidden problems of subsidence along the lines and surrounding areas caused by subway construction and operation,this paper utilizes the time series InSAR technique to obtain the subsidence rate field,cumulative subsidence and time series of subsidence in Nanjing,combines with the precipitation data to study the influence on the subsidence area,and conducts a detailed analysis of the typical subsidence areas of subway lines and their surroundings. The study shows that: Nanjing presents uneven subsidence,located in the Yangtze River diffuse zone,concentrated in the city center areas such as the coast of Jiangning district,Jianye district,Gulou district,etc., in which the subsidence is the most obvious along the central coast of Pukou district,with the highest annual subsidence of -55.9mm,covering an area of about 15.6km²; the subsidence of some lines of Metro Lines 2,4,7,and 10 is more obvious,and in the case of the Tingshan street in Pukou district and the Linjiang-Qingyao Sports park station in Pukou district for example,the highest cumulative settlement of -172.8mm; There is a relationship between Nanjing subway along the rate of settlement and precipitation. Under the impact of heavy rainfall,the pore water in the soil and the pore pressure of the gas contained increases to affect the settlement.

To address the issues of wide lane line distribution range, sparse pixel coverage, and difficulty in feature extraction, this essay constructs a residual lane line detection network based on multi-scale feature fusion. This network is built upon a residual bilateral network and incorporates a bilateral feature aggregation module. It leverages the contextual information from the semantic branch to guide the feature responses of the detail branch within the same stage, thereby integrating information from both branches. Different stages operate at varying scales, and a multi-scale adaptive feature alignment fusion module is used to construct a sampling pre-and post-offset vector index table, reducing detail information loss caused by simple sampling. Additionally, a spatial attention mechanism is introduced to enhance the model's ability to capture long-distance features. Experimental results show that the proposed method performs well across three public datasets, achieving an accuracy of 77.89% on the CULane dataset, which is 2% higher than the current mainstream algorithms.

Deep learning-driven semantic segmentation methods for remote sensing images rely heavily on a large number of manually labeled samples and exhibit poor generalization for unknown tasks, especially in the fine-grained semantic segmentation task where the category system is constantly updated, and the recognition accuracy of the unknown categories (the categories that don't exist in the training samples) needs to be urgently improved. Based on this, the paper proposes a cross-category few-shot segmentation method aimed at multiple farmland categories. The method designs a dual-branch structure, comprising a support branch and a query branch, where the support branch is used for the extraction of segmentation prior, and the query branch is used to complete the propagation of segmentation prior and obtain the segmentation results of the query image. Additionally, the method applies query features to generate self-supporting query prototypes, which significantly improves the expressive ability of the prototypes; a regularization mechanism for prototype alignment between the support and query set is introduced, which makes full use of the knowledge from the support set and improves the discriminative ability of the segmentation. The experiments simultaneously introduce high spatial resolution and hyperspectral image land cover datasets to fully validate the performance of the proposed method. The experimental results show that compared with the existing few-shot segmentation methods, the proposed method can obtain more excellent cross-category farmland recognition results under few-shot conditions.

Aiming at the large format remote sensing image in the chunk boundary feature discontinuity and segmentation efficiency is not high. In this paper, we propose a simple linear iterative clustering superpixel segmentation algorithm (SLIC) that combines the Spark platform and optimal compactness evaluation. Firstly, the SLIC superpixel segmentation method with optimal tightness is used to complete the image chunking, which solves the problem of low accuracy of the chunk boundary; then, the SLIC segmentation algorithm is used in parallel to the chunked data by using Spark to improve the computational efficiency; finally, the SLIC algorithm is improved by using the ratio of vegetation index combined with the method of maximum interclass variation to improve the accuracy of the superpixel segmentation.WorldView-2 Satellite Imagery and GF-2 images are used as experimental data. The experimental results show that the improved SLIC method improves about 9 times of the original method in terms of computing efficiency, 1.5% of the edge fitting precision, 8.2% of the under-segmentation error, and 0.2% of the edge recall.

In mountainous areas with complex terrain, terrain shadows have a great impact on the extraction of remote sensing image information. Therefore, terrain correction should be carried out on remote sensing images to eliminate terrain effects and restore the surface reflectance of terrain shadow areas. This article takes the eastern forest area of Liaoning province (Liaodong Forest Area) as the research area, and uses GF-1 WFV remote sensing images with a spatial resolution of 16m. SCS+C, Minnaert+SCS and SCEDIL correction models are used to perform terrain correction on the original images. Visual analysis, spectral retention effect, terrain correction effect, classification accuracy verification and consistency of spectral reflectance on cloudy and sunny steep slopes are used to compare the images before and after correction, Finally determine the optimal terrain correction model suitable for forest areas. The research results indicate that: ①for forest areas with continuous mountainous and hilly terrain and significant undulations, SCS+C has better spectral retention compared to Minnaert+SCS and SCEDIL models, with a difference of less than 4.32 in the mean reflectance of each band before and after calibration, and there is no overcorrection phenomenon. The terrain correction effect of the three models is judged by the correlation between the corrected near-infrared reflectance and the cosine of the solar incidence angle. The SCS+C model has the smallest correlation, the best terrain correction effect, the Minnaert+SCS model has a slightly larger correlation and the SCEDIL model has overcorrection phenomenon. The image classification accuracy of the SCS+C model after correction has improved by nearly 3% compared to before correction, and is nearly 2% higher than the SCEDIL models of Minnaert+SCS. ②Based on the principle of terrain correction, a new evaluation method for the consistency of spectral reflectance on steep slopes of yin and yang has been added. The impact of NDVI on steep slopes of yin and yang before and after correction is used as the evaluation index for terrain correction effect. The SCS+C correction effect is the best, and the absolute deviation (10^-2) of the mean spectral reflectance on steep slopes of yin and yang before and after correction in two typical areas is reduced from 1.14 to 0.58 and from 1.67 to 0.49, respectively. After correction, the consistency of steep slopes of yin and yang is improved. In summary, the SCS+C model is superior to Minnaert+SCS and SCEDIL, which is more suitable for terrain correction in forest areas.

The ten-parameter calibration model is a classical internal parameter calibration model in near-field photogrammetry. Using this model in image distortion correction, if direct resampling is used, it will produce hollow streaks, while the conventional indirect resampling method does not produce hollows, but it also fails to get better correction effect in the non-central region with large distortion. In order to address this problem, this paper proposes an image Newton iterative accurate correction algorithm for the ten-parameter calibration model. Firstly, the ten-parameter calibration model is used to reorganize the ideal value and actual value of the image point, and construct a system of nonlinear equations with the actual value of the image point as the quantity to be solved; then Newton iteration is used to solve the local linear approximation of the aberrant part of the image; and finally, indirect resampling is used to realize the purpose of accurate correction. Taking the navigation and obstacle avoidance camera of mobile robot as the research object, the experimental results show that the Newton's iterative method of correction can obtain high-precision and high-quality de-distorted images, and the processing effect on the obstacle avoidance camera is especially significant. The algorithm proposed in this paper is suitable for the accurate removal of aberrations from robot's navigation cameras and hazard avoidance cameras, and provides a reliable data source for the generation of high-quality binocular kernel line images for the subsequent stereo matching process.

Aiming at the issues of accuracy and stability being easily affected by NLOS errors in the UWB positioning process, and existing methods of NLOS recognition and removal having complex models and strong environmental dependence, This paper proposes a method for recognizing and removing errors, which combines the a channel characteristic binary hypothesis with an improved UWB autonomous integrity monitoring based on sliding windows. First, this method focuses on identifying and removing significant errors, including gross errors and partial NLOS errors, leveraging the binary assumption of channel characteristics. By doing so, it enhances data quality and reduces the constraints associated with the single ranging fault assumption in UWB localization autonomous integrity monitoring. However, it still faces the challenge of insensitive identification for small-scale NLOS errors. Therefore, by combining the improved UWB localization autonomous integrity monitoring method, which can efficiently identify and remove NLOS ranging values that may not be recognized by the binary assumption method of channel characteristics. Distance measurement accuracy was introduced in the process of constructing inspection factors in autonomous integrity monitoring methods in order to improve accuracy of global test and fault positioning. And the method combines sliding window theory to reduce the missed detection of NLOS, achieving further identification and elimination of NLOS. Experimental results show that the proposed method in this paper is capable of effectively identifying and discarding outlier range measurements containing errors such as NLOS, thereby improving the accuracy and reliability of UWB positioning.

A hierarchical satellite network task planning method for ground observation tasks aimed at monitoring flood disasters in river basins is proposed. Taking the Chongqing and Chenglingji sections of the Yangtze River basin as experimental areas, through passive and active planning. The results show that this method can provide effective multi-scale satellite collaborative observation information for designated time periods of flood disaster areas in the basin. The proposed method can provide important technical support for timely mobilization of satellite resources, rapid response to flood monitoring needs, and acquisition of satellite data for disaster warning, decision-making, and evaluation.

The quality inspection of the reinforced skeleton plays an important role in the production process of prefabricated reinforced concrete components, because its quality directly affects the performance and reliability of the whole component. At present, the quality inspection of rebar skeletons mainly relies on traditional manual means, such as the use of steel ruler measurement and manual counting to evaluate key indicators such as the number and spacing of rebar spacing. However, this approach has significant limitations such as inefficiency,high cost,and error-proneness. In order to solve the above problems,this paper introduces depth camera sensor and advanced technology for the multi-layer reinforcement skeleton in the prefabricated reinforced concrete components,and designs a novel quality inspection method for the reinforcement skeleton,that is,the image is collected by the depth camera,the double-layer reinforcement is converted into a single-layer image according to the depth threshold screening algorithm,and then the intersection coordinates are detected by the improved YOLOv5 algorithm,and finally the binding spacing is solved according to the coordinate conversion. Experimental results show that the proposed method improves the accuracy of inspection,reduces time consumption,and reduces the dependence on manual labor through intelligent algorithms.

Flood monitoring based on network dome camera has become one of the hot spots in water conservancy construction. In order to monitor the flow rate, water level and flow rate in multi-field-of-view, the camera needs to rotate, zoom in and out the focal length. Because of the defects of the camera transmission, the preset point can't be reset accurately when the multi-field-of-view is switched, resulting in the gap error. It creates a large deviation in analyzing water level and flow rate from single image. In this paper, a method of gap error correction for water level measurement is proposed. Firstly, the water level ruler of the preset point is precisely calibrated, and the parameters of the water level equation are obtained. And two correction points with obvious features are selected on the image. Then, The image coordinates of the correction points in the instantaneous field-of-view are linear corrected to the preset field-of-view. Finally, water level in the instantaneous field-of-view is calculated by using the water level equation of the preset point. Through field test, the maximum gap error in different instantaneous field-of-view is 8 pix, and the water level error is 5cm. But the water level error is about 5mm by paper's method. It is shown that the paper's method can greatly reduce gap error in water level measurement.

Human mobility flows are spatio-temporal dependent. The identifying and measuring the spatio-temporal autocorrelation in flows are critical in uncovering human mobility patterns and building prediction models. This study compares and discusses methods of spatial auto-correlation (SFlowLISA) and spatio-temporal auto-correlation (STFlowLISA) to explore spatio-temporal dependencies and aggregation patterns buried in intra-urban and inter-provincial human mobility flows. The results show that:①Spatio-temporal dependencies are significant in both intra-urban and inter-provincial human mobility flows. ②Notably,we observe that flows show high-high (HH) patterns are those short-distance travels,while flows with low-low (LL) patterns are those long-distance travels across regions. ③Specifically,involving both temporal and spatial dependence can effectively capture inter-regional mobility flows than merely measuring spatial dependence. This is particularly important in aggregation patterns of inter-provincial human mobility flows. ④Furthermore,flows with high-low (HL) and low-high (LH) patterns show sharp temporal fluctuation. This characteristic is helpful to identify local outliers in massive flows. Overall,this study emphasizes the advantage and importance of measuring spatio-temporal autocorrelation when analyzing flow data using two typical case studies. The results will benefit the understanding of human mobility patterns and unveiling the auto-correlation characteristics using effective exploratory analysis of STFlowLISA.

Drone tilt photography technology, with its unique advantages of high precision and multi perspective restoration of real landforms, it has been widely applied in fields such as terrain and geomorphology surveying, urban 3D modeling, engineering survey and construction, and land use planning.This study adopts a comprehensive research evaluation method of “unmanned aerial vehicle oblique photography+remote sensing comprehensive interpretation+rockfall trajectory simulation”,based on the 3D slope model of high and steep slopes on both sides of a new highway in Xizang, a detailed interpretation analysis is conducted on the dangerous rock bodies developed in the region, identifying a total of 67 dangerous rock bodies. Through stability analysis and its threat to the road below, it indicates that the dangerous rock mass poses a significant threat to the western central part of the area, the eastern part of the northern slope, and some areas at the foot of the southern slope, which can easily pose a threat to pedestrians and vehicles traveling on the highway.The research results provide important technical basis for the cleaning and protection of hazardous rock masses on site, effectively compensating for the shortcomings of on-site personnels inability to reach and difficult survey operations, and have important theoretical and practical significance.

Aiming at the scale effect of monitoring carbon stock changes based on land cover data, this paper took the example of Hangzhou Chengxi Sci-tech Innovation Corridor to study the applicability of different scales of land cover data in the area with fragmented surface and frequent changes. The coarse and fine classification land cover data obtained from high-resolution images, and the commonly used medium-and low-resolution land cover data, were used to calculate the carbon stock changes in the study area in the past 13 years by using a bookkeeping model. It was found that the value of carbon stock change calculated based on the medium-low resolution land cover decreased by 3021.24tC, compared with that based on the high-resolution coarse classification land cover. And the value of carbon stock change calculated based on the high-resolution coarse classification land cover decreased by 685.43tC, compared with that based on the high-resolution fine classification land cover. The results showed that, in the area of fragmented surface with frequent changes, the carbon stock change values calculated based on different scales of land cover were significantly different, and the high-resolution fine classification land cover data were more applicable.

Mining cracks are a common type of damage that occurs in coal mines due to underground mining. Aiming at the problems of complex surface environment in mining areas and low accuracy of crack extraction methods in UAV images, this studyfused the dynamic snake convolution and dilation convolution to construct a new dynamic snake-dilation convolution. The proposed convolution is added to the encoding and decoding structure of the reference model to optimize the overall network structure; In addition, constructed a crack dataset of the mining area, and verified the accuracy of the crack extraction on this custom dataset. The results show that the addition of dynamic snake-dilation convolution can improve the segmentation accuracy (mean intersection over union) of the model by 14.96%, which is of practical value for achieving accurate extraction of ground cracks.

To improve the weak generalization ability of traditional machine learning methods in remote sensing crop classification, winter wheat classification models that employs Sentinel-2 images with different feature selections and semantic segmentation networks are tested and evaluated in Jiyuan city, Henan province. The results show that compared to the spectral features, the model loss and IoU values of the DeepLab V3+and U-Net++ based on spectral and vegetation indices are reduced and improved by 13.30% and 7.83%, 7.80% and 5.54%, respectively. In addition, the overall accuracy of winter wheat classification results based on U-Net++from 2020 to 2023 is 93.47%～95.60%, which is 0.12%～2.29% and 4.84%～7.40% higher than that of DeepLab V3+ and random forest, respectively. Moreover, the landscape metrics values also indicate that the winter wheat classification results based on U-Net++network perform better patch integrity and compactness. Finally, the change data and spatial distribution of winter wheat based on U-Net++ from 2020 to 2023 are analyzed. It can provide methodological support for practical applications such as crop area monitoring under complex terrain conditions.

In order to solve the problem that manned helicopters and compound wing UAVs in high altitude and mountainous areas operate at high relative altitude, are vulnerable to crosswind and have sparse point clouds, this paper proposes a method to obtain three-dimensional laser point cloud data of overhead transmission lines by “tower imitation flight”,optimizes the traditional “ground imitation flight” mode, and obtains point cloud data of a 500kV ultra-high voltage line of Shannan in Tibet.After data processing and analysis, the obtained point cloud data has a high voxel density and high spatial positional accuracy. The experimental results show that the high-capacity multi rotor aircraft conducting “tower imitation flight” operations in high-altitude and mountainous areas not only reduces operational risks, but also obtains high point cloud data density, low noise, and obvious micro characteristics of hardware and wires. At the same time, this method has the advantages of high maneuverability, strong stability, and low operational risks, providing support for the fine autonomous inspection of overhead transmission lines and the intelligent application of the State Grid.

InSAR technology has the advantages of large range, high precision, all-day and all-weather. It has been widely used in surface deformation monitoring and elevation acquisition. The GAMMA software has the advantages of supporting rich data sources, comprehensive processing functions, reliable result accuracy and high efficiency. GAMMA is currently a commonly used InSAR data processing software in the world. However, GAMMA uses a command line user interface for interaction, which leads its complicated operating and high professionalism requirements. So it is difficult for non-professional users to use. In order to solve the above problems, this paper uses the user-friendly and open source QGIS to carry out secondary development of GAMMA, combines the py＿gamma library provided by GAMMA and the py_QGIS API interface provided by QGIS, taking the development of InSAR deformation monitoring system as a case, and taking the 106 Sentinel-1 in Warda area,Xizang No. is processing the experiment, which proves the feasibility and reliability of secondary development based on QGIS and GAMMA in terms of software performance and processing results. At the same time, the development adopts technologies such as modular design and algorithm parallel optimization. Relevant content can provide reference for converting command line user interface software to graphical user interface.

With the proposal of carbon neutrality goals, the development and utilization of new energy are growing. Due to the numerous factors involved in project design, it is difficult to select the location of new energy sources under the constraints of land and geography. Facing the explosive growth of new energy projects, traditional site selection design methods urgently need to improve efficiency to ensure the scientific and rational nature of new energy projects. Based on the analysis of the principles and application status of GIS+BIM technology, this article proposes a new solution for intelligent site selection of new energy sources. A 3D digital platform is constructed based on the integration of GIS+BIM technology. Through the platform, various resource conditions such as geographical location and environmental factors are fully analyzed, and comprehensive evaluation and optimization of new energy project locations are achieved. Through application analysis, the research results of the article can ensure the scientific and rationality of site selection, improve design efficiency by 15%, improve site selection accuracy, reduce costs, and promote the sustainable development of new energy projects.

As a new form of map industry, smart driving high-precision maps contain high-rich, high-precision and high-freshness geographical information data, which are related to national sovereignty, security and interests. The current high-precision map review work lacks effective automated review technology, standard databases and institutional guarantees, making it difficult to meet the demand for map freshness by smart driving, which will have a certain impact on smart driving. This article summarizes the development status of digital review of high-precision maps for smart driving, sorts out relevant policy trends, systematically analyzes the key technologies of digital map review and the difficulties in large-scale application, and combines the map review business model to propose solutions for the digital review technology route for high-precision maps for smart driving aims to provide a reference for efficient and reliable high-precision map review.

Steel structures will produce spatial deformation under long-term loading and uneven forces, and deflection is an important measure of it. Conventional deflection detection methods rely on manual measurements or fixed sensors, which have certain limitations in real-time, accuracy and coverage, making it difficult to accurately capture the subtle changes in structural deformation. In this paper, the plant project of an automobile is taken as the engineering background, and the 3D laser scanning system based on simultaneous localization and mapping (SLAM) technology is used to carry out jacking measurements on the steel structure mesh frame and extract the spherical centers of the node balls of the steel structure through the RANSAC algorithm for deflection detection. The study shows that the steel structure deflection detection method based on SLAM technology has high accuracy and real-time performance, and can effectively detect the deflection deformation of the steel structure mesh frame, which provides a new idea and solution for steel structure deflection detection, and has important theoretical and application value.