An ecological red line dynamic monitoring method and system based on remote sensing images

Abstract

The application relates to the technical field of intelligent monitoring, and discloses an ecological red line dynamic monitoring method and system based on remote sensing images, the method comprising the following steps: performing data preprocessing on multi-temporal remote sensing image data of a target region to obtain a standard image sequence; performing ecological element characterization extraction on the standard image sequence to obtain spectral features and spatial texture features; performing edge optimization segmentation on the standard image sequence to obtain an ecological red line region vector diagram; superimposing and comparing continuous ecological red line region vector diagrams to identify spatial distribution information of the ecological red line region vector diagrams, so as to obtain a change region patch set; performing feature change mode determination on the change region patch set to obtain natural evolution states and human activity interference information; and integrating the spatial distribution information, the natural evolution states and the human activity interference information into an ecological red line dynamic monitoring report; and the application can improve the efficiency of ecological red line dynamic monitoring based on remote sensing images.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and in particular to a method and system for dynamic monitoring of ecological red lines based on remote sensing images. Background Technology

[0002] In the field of dynamic monitoring of ecological red lines, remote sensing image-based monitoring methods have been applied to some extent, but existing technologies have significant limitations in data preprocessing. Current technologies often struggle to comprehensively and accurately standardize multi-temporal remote sensing image data. Insufficient coordination in radiometric calibration, atmospheric correction, and geometric deformation correction leads to deviations in spatial alignment and boundary uniformity between images from different time phases. This limits the accuracy of subsequently extracted monitoring baseline data and fails to provide reliable standard image support for ecological red line monitoring. Furthermore, existing technologies extract relatively singular dimensions of ecological element representation, failing to fully integrate spectral and spatial texture features. This results in an incomplete characterization of the ecological status of the target area, affecting the scientific rigor of subsequent monitoring and analysis.

[0003] Existing technologies are inadequate in delineating ecological red line areas and identifying dynamic changes. During edge segmentation, existing technologies lack refined optimization methods for initial boundaries and fail to fully integrate ecological prior knowledge and semantic context information to construct effective constraint mechanisms, resulting in insufficient accuracy in ecological red line boundary delineation and difficulty in ensuring topological consistency. Furthermore, in the identification and feature determination stages of change areas, existing technologies lack efficient spatial topological analysis methods for overlaying and comparing continuous temporal vector maps, and the selection criteria for candidate change areas are not reasonable enough, easily generating invalid interference information. Simultaneously, the determination of change patterns lacks systematic rule base support, making it difficult to accurately distinguish between the driving types of natural evolution and human-induced disturbances. Ultimately, this leads to insufficient reliability of monitoring results, and the data integration and report generation processes are cumbersome and inefficient, failing to meet the real-time and accuracy requirements of dynamic monitoring of ecological red lines. Summary of the Invention

[0004] This invention provides a method and system for dynamic monitoring of ecological red lines based on remote sensing images, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a method for dynamic monitoring of ecological red lines based on remote sensing images, comprising:

[0006] S1. Perform data preprocessing on the multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area;

[0007] S2. Extract ecological element characteristics from the standard image sequence to obtain the spectral and spatial texture features of the target region;

[0008] S3. Based on the spectral features and spatial texture features, perform edge optimization segmentation on the standard image sequence to obtain the ecological red line region vector map of the target region;

[0009] S4. In the target area, the continuous ecological red line area vector maps are superimposed and compared to identify the spatial distribution information of the ecological red line area vector maps, so as to obtain the set of change area patches in the target area.

[0010] S5. Determine the characteristic change pattern of the patch set in the changed area to obtain the natural evolution state and human activity interference information of the target area;

[0011] S6. Integrate the spatial distribution information, the natural evolution status, and the human activity interference information into a dynamic monitoring report of the ecological red line of the target area.

[0012] In a preferred embodiment, the step of preprocessing the multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area includes:

[0013] Radiometric calibration is performed on multi-temporal remote sensing image data of the target area to obtain radiometric image data of the multi-temporal remote sensing image data;

[0014] Atmospheric correction is performed on the radiance image data to obtain the surface reflectance image of the target area;

[0015] Based on the preset digital elevation model and ground control points in the target area, the geometric deformation in the surface reflectance image is corrected to obtain the orthorectified image data of the target area.

[0016] The orthorectified images are registered and cropped to align and unify the orthorectified images from different time phases to the boundary of the target region, thus obtaining a standard image sequence of the target region.

[0017] In a preferred embodiment, the step of extracting ecological element characteristics from the standard image sequence to obtain the spectral and spatial texture features of the target region includes:

[0018] Super-resolution reconstruction is performed on the standard image sequence to obtain enhanced image data of the target region;

[0019] Based on the multi-band reflectance data in the standard image sequence, statistical analysis is performed on the land cover type index of the target area to obtain the spectral characteristics of the target area.

[0020] The enhanced image data is analyzed for grayscale spatial relationship features to obtain the spatial texture features of the target region.

[0021] In a preferred embodiment, the step of performing edge optimization segmentation on the standard image sequence based on the spectral features and spatial texture features to obtain the ecological redline region vector map of the target region includes:

[0022] The pixel-level semantic features in the standard image sequence are subjected to full-image scanning classification to obtain a preliminary land cover classification map of the target area.

[0023] Edge detection is performed on the preliminary land feature classification map to obtain the preliminary boundary pixels of the target area;

[0024] Based on the semantic category and spatial proximity of the preliminary boundary pixels, the preliminary land cover classification map is subjected to region growing and merging to obtain candidate ecological redline patches for the target area;

[0025] The candidate ecological redline patches are finely optimized to obtain the ecological redline boundary of the target area;

[0026] Based on the topological consistency of the target area, a geographic vector transformation is performed on the ecological red line boundary to obtain a vector map of the ecological red line area of ​​the target area.

[0027] In a preferred embodiment, the step of finely optimizing the candidate ecological redline patches to obtain the ecological redline boundary of the target area includes:

[0028] Multi-scale deep feature extraction is performed on the surrounding area of ​​the candidate ecological red line patch to obtain the semantic context information of the preliminary boundary pixels;

[0029] Based on a pre-defined ecological prior knowledge base, the energy function of the target region is constructed using the coordinates of the boundary points of the initial boundary in the candidate ecological red line patch as variables, and the smoothness constraint term of the boundary continuity in the ecological prior knowledge base and the semantic context information as common constraints.

[0030] The candidate ecological redline patches are optimized by minimizing graph cut using the energy function to obtain the lowest energy boundary point of the candidate ecological redline patches.

[0031] The ecological red line boundary of the target area is determined based on the minimum energy boundary point.

[0032] In a preferred embodiment, the formula for calculating the relevant energy boundary points in the energy function is as follows:

[0033] ;

[0034] In the formula, The set of candidate boundary points in the candidate ecological redline patches The overall energy value, The preset semantic balance weight coefficients, Boundary points in the initial boundary The degree of consistency of semantic context information The preset adaptive smoothness weight coefficients, For semantic context confidence, For the smoothness constraint term of the ecological red line boundary, Boundary point The neighborhood pixel set.

[0035] In a preferred embodiment, the step of overlaying and comparing consecutive ecological redline area vector maps in the target area to identify the spatial distribution information of the ecological redline area vector maps, thereby obtaining a set of change area patches in the target area, includes:

[0036] The vector maps of the ecological red line areas in two adjacent time phases are superimposed and projected in the same spatial coordinate system to obtain a vector superimposed layer of the target area;

[0037] Spatial topological analysis is performed on the vector overlay layer to obtain a set of candidate change regions for the vector overlay layer;

[0038] Based on the preset change area threshold and change shape rules, the candidate change region set is filtered and eliminated to obtain the effective change region set of the target region;

[0039] The effective change areas are concentrated into independent regions, and the spatial location and boundaries of the independent regions are encapsulated as independent geographic features to obtain the change area patch set of the target area.

[0040] In a preferred embodiment, the step of determining the characteristic change pattern of the patch set in the changed region to obtain the natural evolution state and human activity interference information of the target region includes:

[0041] Key element analysis was performed on the individual changed patches in the changed area to obtain the dominant ecological element type, spatial geometry and temporal span characteristics of the changes in the individual changed patches;

[0042] The dominant ecological element type, spatial geometric morphology, and temporal span characteristics are retrieved and evaluated against a preset natural evolution discrimination rule base and an anthropogenic interference discrimination rule base to obtain the driving type of the changed patch.

[0043] When the driving type is the natural evolution mode, the natural evolution state of the target area is obtained by synchronously analyzing the conversion direction and spatial aggregation degree of the dominant ecological element type.

[0044] When the driving type is the human interference mode, the human activity interference information of the target area is obtained by coordinating the transformation mode of the dominant ecological element type and the spatial distribution intensity.

[0045] In a preferred embodiment, integrating the spatial distribution information, the natural evolution status, and the information on human disturbance into a dynamic monitoring report of the ecological red line for the target area includes:

[0046] By combining the spatial distribution information of the changed patch set, the natural evolution status information, and the human activity interference information, a multi-source dynamic monitoring dataset of the target area is obtained.

[0047] The data in the multi-source dynamic monitoring dataset are unified in terms of spatiotemporal reference, and spatial correlation analysis is performed on the unified dataset to obtain a structured spatiotemporal database of the target area.

[0048] Based on the structured spatiotemporal database and the preset report template framework, the spatial distribution information and the human activity interference information are spatially overlaid and visualized.

[0049] By associating and evaluating the comprehensive information after spatial overlay and visualization rendering with the text describing the natural evolution state, a comprehensive map and a draft of the analysis text of the monitoring results of the target area are obtained.

[0050] The completeness and logical consistency of the comprehensive map and the initial draft of the analysis text of the monitoring results are verified and integrated to obtain the dynamic monitoring report of the ecological red line of the target area.

[0051] To address the aforementioned problems, the present invention also provides a dynamic monitoring system for ecological red lines based on remote sensing images, the system comprising:

[0052] The data preprocessing and input module is used to preprocess multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area.

[0053] The feature extraction module is used to extract ecological element characteristics from the standard image sequence to obtain the spectral features and spatial texture features of the target region;

[0054] The change detection module is used to perform edge optimization segmentation on the standard image sequence based on the spectral features and spatial texture features to obtain the ecological red line region vector map of the target region;

[0055] The change area vectorization module is used to overlay and compare continuous ecological red line area vector maps in the target area to identify the spatial distribution information of the ecological red line area vector maps, so as to obtain the change area patch set of the target area;

[0056] The change pattern determination module is used to determine the feature change patterns of the patch set in the changed region to obtain the natural evolution state and human activity interference information of the target region.

[0057] The results integration and output module is used to integrate the spatial distribution information, the natural evolution status, and the human activity interference information into a dynamic monitoring report of the ecological red line of the target area.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. This invention employs a standardized preprocessing workflow, including radiometric calibration, atmospheric correction, geometric deformation correction, and registration / cropping, for multi-temporal remote sensing image data. This ensures spatial alignment and boundary uniformity across images from different time periods, constructing a high-quality standard image sequence. Simultaneously, super-resolution reconstruction technology enhances the image data, and spectral and spatial texture features are extracted concurrently. This achieves multi-dimensional coverage and in-depth analysis of ecological elements, significantly improving the comprehensiveness and accuracy of the ecological status characterization of the target area, and providing solid technical support for subsequent ecological redline monitoring.

[0060] 2. This invention constructs an energy function that integrates semantic context information and ecological smoothness constraints, and achieves precise delineation of ecological red line boundaries through minimum graph cut optimization, significantly improving the accuracy and topological consistency of red line boundaries. In the change monitoring and results integration stage, effective change areas are accurately screened through spatial topological analysis, and the precise determination of natural evolution and human interference is completed based on a dedicated rule base. A complete monitoring report is then generated through spatiotemporal benchmark unification, spatial correlation analysis, and visualization rendering. This not only improves the efficiency of dynamic monitoring of ecological red lines but also ensures the systematic nature and practicality of monitoring results, providing comprehensive and reliable information support for the dynamic management and scientific decision-making of ecological red lines. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a method for dynamic monitoring of ecological red lines based on remote sensing images, provided in an embodiment of the present invention. Figure 2 A functional block diagram of an ecological red line dynamic monitoring system based on remote sensing images provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0062] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0063] This application provides a method for dynamic monitoring of ecological red lines based on remote sensing images. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for dynamic monitoring of ecological red lines based on remote sensing images can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0064] Reference Figure 1 The diagram shown is a flowchart illustrating a method for dynamic monitoring of ecological red lines based on remote sensing images, according to an embodiment of the present invention. In this embodiment, the method for dynamic monitoring of ecological red lines based on remote sensing images includes:

[0065] S1. Perform data preprocessing on the multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area;

[0066] In this embodiment of the invention, the step of preprocessing the multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area includes:

[0067] Radiometric calibration is performed on multi-temporal remote sensing image data of the target area to obtain radiometric image data of the multi-temporal remote sensing image data;

[0068] Atmospheric correction is performed on the radiance image data to obtain the surface reflectance image of the target area;

[0069] Based on the preset digital elevation model and ground control points in the target area, the geometric deformation in the surface reflectance image is corrected to obtain the orthorectified image data of the target area.

[0070] The orthorectified images are registered and cropped to align and unify the orthorectified images from different time phases to the boundary of the target region, thus obtaining a standard image sequence of the target region.

[0071] When performing radiometric calibration on multi-temporal remote sensing image data of the target area, the digital quantization value corresponding to each pixel in the remote sensing image is extracted pixel by pixel. Combined with the radiometric calibration coefficients set by the sensor at the factory, the digital quantization value of each pixel is linearly transformed to eliminate the error caused by the non-uniformity of the sensor's own radiometric response. The digital quantization value representing the image brightness is converted into the actual radiometric physical quantity of the corresponding ground object. Finally, radiometric image data of multi-temporal remote sensing image data with radiometric brightness as the measurement index for each pixel is obtained.

[0072] When performing atmospheric correction processing on radiance image data, the radiance image data is used as the basic processing object. Combined with the atmospheric parameters at the time of the remote sensing image of the target area, which are obtained through atmospheric sounding data of the same period in the target area, the atmospheric parameters include atmospheric aerosol type, atmospheric water vapor content, and atmospheric ozone concentration. For the radiance value of each pixel in the radiance image data, the radiation attenuation error caused by atmospheric scattering, atmospheric absorption, and atmospheric radiation is eliminated in sequence. The radiance value after atmospheric attenuation is restored to the physical quantity of the surface reflectance of the ground object itself, and finally the surface reflectance image of the target area is obtained.

[0073] When performing geometric deformation correction processing based on the preset digital elevation model and ground control points in the target area, the ground control points used are feature points evenly distributed within the target area, with no fewer than 50 feature points, and each ground control point has accurate latitude and longitude geographic coordinates and elevation values. The digital elevation model used is 30-meter resolution raster elevation data within the target area. Taking the surface reflectance image as the processing object, the image coordinates corresponding to each pixel in the image are first extracted, and then the latitude and longitude geographic coordinates of the ground control points are accurately matched with the image coordinates. Combined with the elevation data provided by the digital elevation model, the pixel position offset problem caused by terrain undulation and sensor imaging side view in the surface reflectance image is corrected pixel by pixel coordinates. The deviation of each pixel from the actual geographic spatial position is corrected, and finally the orthorectified image data of the target area is obtained.

[0074] When performing image registration and cropping on orthorectified images, the orthorectified image data is used as the processing object. First, an orthorectified image of the target area that is free from cloud cover and band loss during imaging is selected as the registration reference image. Orthorectified images of all other time phases are registered with this reference image. During the registration process, the pixel spatial resolution of the reference image is used as a unified standard. Pixel resampling is performed on the orthorectified images of other time phases to ensure that the pixel spatial resolution of orthorectified images of all time phases remains consistent. Then, based on the preset boundary latitude and longitude coordinate range of the target area, cropping is performed on all registered orthorectified images of all time phases. During the cropping process, pixels within the corresponding range are strictly extracted according to the preset boundary latitude and longitude coordinate range, and all pixel data outside the coordinate range are removed. This ensures that the orthorectified images of different time phases are completely aligned in geographic space, and the coverage of all images accurately matches the boundary range of the target area, ultimately obtaining a standard image sequence of the target area.

[0075] The beneficial effects include comprehensive preprocessing of multi-temporal remote sensing image data of the target area, successively eliminating various errors such as sensor radiometric response inhomogeneity, radiometric attenuation caused by atmospheric scattering and absorption and radiation, and geometric offset caused by topographic undulation and sensor imaging side view. This accurately restores the actual radiance and surface reflectance of the ground features, achieving precise matching between remote sensing image pixel positions and actual geographic spatial positions. At the same time, it ensures that remote sensing images from different time phases are completely aligned in geographic space and that the coverage is accurately unified to the boundary of the target area. The resulting standard image sequence has accurate physical quantity representation of ground features and a unified spatial benchmark, providing high-quality, comparable, and reproducible basic image data for subsequent analysis and processing work based on multi-temporal remote sensing images.

[0076] S2. Extract ecological element characteristics from the standard image sequence to obtain the spectral and spatial texture features of the target region;

[0077] In this embodiment of the invention, the step of extracting ecological element characteristics from the standard image sequence to obtain the spectral features and spatial texture features of the target region includes:

[0078] Super-resolution reconstruction is performed on the standard image sequence to obtain enhanced image data of the target region;

[0079] Based on the multi-band reflectance data in the standard image sequence, statistical analysis is performed on the land cover type index of the target area to obtain the spectral characteristics of the target area.

[0080] The enhanced image data is analyzed for grayscale spatial relationship features to obtain the spatial texture features of the target region.

[0081] When performing super-resolution reconstruction on a standard image sequence, the reflectance values ​​of each pixel in the standard image sequence are extracted pixel by pixel. The correlation characteristics and spatial distribution patterns of the reflectance values ​​of each pixel with its neighboring pixels are analyzed. Based on these patterns, the spatial detail information of each pixel is completed and restored, filling in the missing details such as the edges of ground features and minute textures in the standard image sequence due to spatial resolution limitations. This improves the spatial resolution of the image, making the morphological features and spatial distribution details of various ground features in the image clearer, and finally obtaining the enhanced image data of the target area.

[0082] When conducting statistical analysis of land cover type index based on multi-band reflectance data from standard image sequences, all band reflectance data of each temporal image in the standard image sequence are first extracted. Independent statistical analysis units are then divided according to the preset land cover type in the target area. Each analysis unit corresponds to a land cover type. The reflectance data of each band for all pixels in each analysis unit are summarized. The corresponding land cover type index is calculated and statistically analyzed for each land cover type. The distribution patterns and numerical characteristics of the land cover type index corresponding to different land cover types are identified and determined. These patterns and characteristics are then integrated as the spectral characteristics of the target area.

[0083] When performing grayscale spatial relationship feature analysis on enhanced image data, the grayscale values ​​of all pixels in the enhanced image data are extracted pixel by pixel. The grayscale value correlation between adjacent pixels, the overall distribution of grayscale values ​​of pixels within the same land feature area, and the grayscale value change characteristics at the junction of different land features are systematically analyzed. At the same time, the spatial distribution patterns and grayscale value correlation characteristics of various typical land features in the target area are sorted out. The integrated grayscale spatial correlation patterns and distribution characteristics of various land features are used as the spatial texture features of the target area.

[0084] The beneficial effects are as follows: enhanced image data is obtained by super-resolution reconstruction of standard image sequences; spectral features of the target area are obtained by statistical analysis of multi-band reflectance data in the standard image sequences; spatial texture features of the target area are obtained by gray-scale spatial relationship feature analysis of the enhanced image data; the spatial detail rendering capability of the image is effectively improved; spectral and texture-related feature information of ground objects in the target area is accurately captured; and ecological element representation of the target area is fully obtained, providing accurate and comprehensive feature support for subsequent analysis and processing work based on ecological elements.

[0085] S3. Based on the spectral features and spatial texture features, perform edge optimization segmentation on the standard image sequence to obtain the ecological red line region vector map of the target region;

[0086] In this embodiment of the invention, the step of performing edge optimization segmentation on the standard image sequence based on the spectral features and spatial texture features to obtain the ecological redline region vector map of the target region includes:

[0087] The pixel-level semantic features in the standard image sequence are subjected to full-image scanning classification to obtain a preliminary land cover classification map of the target area.

[0088] Edge detection is performed on the preliminary land feature classification map to obtain the preliminary boundary pixels of the target area;

[0089] Based on the semantic category and spatial proximity of the preliminary boundary pixels, the preliminary land cover classification map is subjected to region growing and merging to obtain candidate ecological redline patches for the target area;

[0090] The candidate ecological redline patches are finely optimized to obtain the ecological redline boundary of the target area;

[0091] Based on the topological consistency of the target area, a geographic vector transformation is performed on the ecological red line boundary to obtain a vector map of the ecological red line area of ​​the target area.

[0092] The process of finely optimizing the candidate ecological redline patches to obtain the ecological redline boundary of the target area includes:

[0093] Multi-scale deep feature extraction is performed on the surrounding area of ​​the candidate ecological red line patch to obtain the semantic context information of the preliminary boundary pixels;

[0094] Based on a pre-defined ecological prior knowledge base, the energy function of the target region is constructed using the coordinates of the boundary points of the initial boundary in the candidate ecological red line patch as variables, and the smoothness constraint term of the boundary continuity in the ecological prior knowledge base and the semantic context information as common constraints.

[0095] The candidate ecological redline patches are optimized by minimizing graph cut using the energy function to obtain the lowest energy boundary point of the candidate ecological redline patches.

[0096] The ecological red line boundary of the target area is determined based on the minimum energy boundary point.

[0097] The formulas for calculating the relevant energy boundary points in the energy function are as follows:

[0098] ;

[0099] In the formula, The set of candidate boundary points in the candidate ecological redline patches The overall energy value, The preset semantic balance weight coefficients, Boundary points in the initial boundary The degree of consistency of semantic context information The preset adaptive smoothness weight coefficients, For semantic context confidence, For the smoothness constraint term of the ecological red line boundary, Boundary point The neighborhood pixel set.

[0100] The candidate boundary point set is derived from the set of all boundary points on the initial boundary of the candidate ecological redline patch. The initial boundary is determined by the spatial outline boundary delineated by the candidate ecological redline patch after regional growth and merging. The set contains the spatial coordinates and corresponding semantic category information of each boundary point on the initial boundary.

[0101] The preset semantic balance weight coefficient is derived from the quantitative setting of the influence of semantic context information during the optimization of ecological red line boundaries. This setting is based on the historical data of ecological red line delineation in the target area. The specific value is determined after statistically analyzing the proportion of the influence of semantic context information on boundary point selection in the historical data, so as to ensure that the coefficient can accurately match the ecological land cover distribution characteristics of the target area.

[0102] The degree of agreement between the semantic context information of the boundary points comes from the quantitative result of matching the semantic context information extracted from the multi-scale deep features corresponding to the boundary points with the feature information of the ecological red line land cover categories in the preset ecological prior knowledge base. During the matching process, the overlap of feature information is used as the basis to numerically represent the agreement between the semantic information of the boundary points and the prior features.

[0103] The preset adaptive smoothness weight coefficient is derived from the quantitative setting of the influence of smoothness constraint terms in the ecological red line boundary optimization process. This setting combines the morphological characteristics of the natural continuity of ecological boundaries in the ecological prior knowledge base, and refers to the boundary smoothness characteristics of different ecological land cover types in the target area. After classification and statistics, the specific value is determined so that the coefficient can adapt to the boundary smoothness constraint requirements of different land cover types.

[0104] Semantic context confidence is derived from a quantitative assessment of the effectiveness and reliability of the semantic context information corresponding to the boundary point. The assessment is based on the completeness of multi-scale deep feature extraction in the area surrounding the boundary point. The more complete the extracted feature information, the higher the confidence value. This assessment standard is based on the technical specifications of multi-scale feature extraction to ensure that the assessment results can truly reflect the quality of semantic context information.

[0105] The ecological red line boundary smoothness constraint term is derived from the quantitative representation result of combining the smoothness constraint requirements of the ecological red line boundary continuity in the ecological prior knowledge base with the features of the neighboring pixel set of the boundary point. The determination of this result is based on the spatial positional relationship between the boundary point and the pixels in the neighboring pixel set, and is formed by matching the morphological features of the natural continuity of the boundary in the ecological prior knowledge base.

[0106] The neighborhood pixel set of the boundary point is derived from the set of surrounding pixels within a fixed pixel range centered on the boundary point. This pixel range is determined based on the scale setting of multi-scale depth feature extraction of the surrounding area of ​​the candidate ecological redline patch. The set contains the spectral features, spatial texture features and land cover category information of each neighboring pixel.

[0107] This formula is the core expression for quantifying the overall energy value of the candidate boundary point set in the candidate ecological redline patch. By integrating the dual constraints of semantic and smoothness levels, it transforms the dual screening requirements for boundary points in the ecological redline boundary optimization process into quantifiable numerical results. This provides a direct numerical basis for minimizing the graph cut optimization of candidate ecological redline patches. The formula uses two types of weight coefficients to adjust the influence of semantic context information and smoothness constraints on the overall energy value calculation, ensuring that the calculation results accurately match the actual needs of ecological redline delineation in the target area. At the same time, the formula combines the degree of consistency and confidence of semantic context information to ensure the accuracy and reliability of semantic constraint judgment. The ecological redline boundary smoothness constraint ensures the constraint requirement of natural continuity of the boundary at the morphological level. The final calculated overall energy value can intuitively reflect whether the candidate boundary point set simultaneously meets the dual constraints of semantic and smoothness. The lower the overall energy value, the more the corresponding candidate boundary point set conforms to the optimization standard of the ecological redline boundary. This provides an accurate and reproducible numerical judgment standard for subsequently screening the lowest energy boundary points from the candidate boundary point set and determining the ecological redline boundary of the target area.

[0108] Based on the pixel row and column arrangement order of the standard image sequence, starting from the first pixel in the upper left corner of the image, the entire image is scanned row by row and column by column until the last pixel in the lower right corner. During the scanning process, the corresponding spectral features and spatial texture features are extracted for each pixel to form the pixel-level semantic features of that pixel. The extracted pixel-level semantic features are matched one by one with the features of each category in the preset land cover category feature library. When the feature overlap reaches 95% or more, the pixel is directly classified into the corresponding land cover category. Pixels that do not reach this threshold are combined with the pixel-level semantic features of their four neighboring pixels above, below, left, and right for secondary feature matching. The secondary matching is still based on the 95% feature overlap as the judgment criterion to determine their land cover category. After the land cover category determination of all pixels in the standard image sequence is completed, pixels of the same category are spatially aggregated according to the original spatial position of each pixel in the standard image sequence to form a preliminary land cover classification map of the target area with clear land cover category division and category boundaries.

[0109] For each pixel in the preliminary land cover classification map, the land cover category information of that pixel is extracted as the center pixel information. At the same time, the land cover category information of the eight neighboring pixels above, below, left, right, front, back, and back of that pixel is extracted. The land cover category of the center pixel is compared with the land cover categories of the eight neighboring pixels one by one. When the land cover category of the center pixel is inconsistent with any of the neighboring pixels, the center pixel is marked as a boundary candidate pixel. After completing the marking operation of all pixels in the preliminary land cover classification map, spatial filtering is performed on all boundary candidate pixels. The filtering condition is to retain only the boundary candidate pixels that have three or more adjacencies with pixels of different types of land cover. The pixels that have passed the filtering condition are determined as preliminary boundary pixels. The set of all preliminary boundary pixels is the preliminary boundary pixels of the target area, and each preliminary boundary pixel retains its original spatial coordinate information in the preliminary land cover classification map.

[0110] The semantic category of each preliminary boundary pixel is extracted. Simultaneously, using the spatial coordinates of the pixels as the criterion, if the difference in horizontal or vertical coordinates between two preliminary boundary pixels does not exceed one pixel, the two preliminary boundary pixels are considered to have a spatial proximity relationship. After completing the semantic category extraction and spatial proximity analysis of all preliminary boundary pixels, each independent land cover category region in the preliminary land cover classification map is used as the initial seed region for region growth. The preliminary boundary pixels serve as the boundary reference for region growth. Growth and extension begin from the edge pixels of the seed region and extend towards its adjacent pixel regions. The criterion for growth and extension is the number of land cover categories to be merged. If the semantic categories of the land cover region and the seed region are consistent and the number of initial boundary pixels between the two regions is 0, then the adjacent land cover regions of the same category that meet this condition will be spatially merged with the seed region. During the merging process, redundant internal boundaries between the two regions will be eliminated simultaneously. After completing the region growth and merging operation of all land cover regions that meet the conditions in the preliminary land cover classification map, land cover regions that meet the criteria will be extracted from all the merged land cover regions according to the land cover category standards defined by the ecological red line. According to the pixel spatial distribution of the extracted land cover regions, their spatial contours will be delineated to form candidate ecological red line patches of the target region with clear spatial range and contour boundaries.

[0111] For each pixel on the outline boundary of a candidate ecological redline patch, the spectral and spatial texture features corresponding to that pixel are re-extracted by combining the spectral and spatial texture features of the standard image sequence. Simultaneously, the spectral and spatial texture features of all pixels within a three-pixel radius surrounding that pixel are extracted. The feature information of the boundary pixel is then fused with the feature information of the pixels within the three-pixel radius. Based on the feature criteria of ecological redline land cover categories, the land cover category of the boundary pixel is re-determined. If the re-determination result is a non-ecological redline land cover category, the pixel is removed from the outline boundary of the candidate ecological redline patch; otherwise, it is removed from the boundary. For redline land cover categories, the pixel is retained in the outline boundary of the candidate ecological redline patch. After re-judging and removing all boundary pixels, for the empty positions in the outline boundary caused by pixel removal, the ecological redline land cover category pixels within a range of 3 pixels around the empty position are used as a reference to complete the outline boundary. The selected pixels for completion must have a feature overlap of 95% or more with the feature standard of the ecological redline land cover category. After completing the removal and completion of the outline boundaries of all candidate ecological redline patches, the optimized outline boundary pixels are connected continuously in spatial order according to their spatial coordinates to form a continuous and uninterrupted ecological redline boundary of the target area.

[0112] The topological consistency criteria for the target area are defined as follows: the spatial contours of the ecological redline boundaries must be free of self-intersection, spatial overlap, and contour breaks. Based on this standard, the topological consistency of the ecological redline boundaries is verified. During the verification process, the spatial connection relationships between boundary segments are checked segment by segment according to the pixel spatial coordinates of the ecological redline boundaries. If a spatial self-intersection is detected, redundant boundary pixels at that location are removed, and the contours are reconnected according to the ecological redline land cover category feature standards. If a spatial overlap is detected, boundary pixels in that area that meet the ecological redline land cover category standards are retained, and duplicate contours in the overlapping parts are eliminated. If a contour break is detected, the ecological redline boundary pixels on both sides of the break are used as references to supplement the contours with the ecological redline land cover category feature standards. After completing the connection and repair of contour breakage locations for pixels with a feature overlap of 95% or higher, and verifying the topological consistency of the ecological redline boundary and correcting all issues, the pixel plane coordinates of each pixel on the ecological redline boundary are converted into corresponding geographic latitude and longitude coordinates using the geographic coordinate system corresponding to the target area as the conversion benchmark. Following the spatial continuity order of the ecological redline boundary, all the converted geographic latitude and longitude coordinates are connected sequentially to form geographic vector elements composed of geographic coordinate points, geographic contour lines, and geographic areas. In accordance with the organization specifications of geospatial data, all the formed geographic vector elements are systematically organized, and each geographic vector element is assigned corresponding ecological redline attribute information, ultimately forming an ecological redline area vector map of the target area that conforms to the geospatial data specifications.

[0113] Taking the initial boundary of the candidate ecological redline patch as the core, the preliminary boundary pixel is used as the feature extraction center. The neighborhood range of 3×3, 5×5 and 7×7 pixels are used as different scales to extract features from the neighborhood region corresponding to each preliminary boundary pixel at each scale. During the extraction process, the spectral features and spatial texture features of each scale region are acquired simultaneously. Then, the features extracted from the same preliminary boundary pixel at different scales are fused layer by layer. During the fusion, the core information related to the ecological redline land cover category in the features of each scale is retained. Finally, the fused feature information is uniquely associated with the corresponding preliminary boundary pixel. The information set containing the multi-scale fused features and pixel association relationship is the semantic context information of the preliminary boundary pixel.

[0114] The pre-defined ecological prior knowledge base includes the boundary spatial distribution patterns of typical land features in ecological redline areas, semantic association rules of land features, and the morphological characteristics of natural continuity of ecological boundaries. The spatial coordinates of each boundary point in the initial boundary of the candidate ecological redline patches are incorporated into the function construction system as adjustable variables. The smoothness constraint term for boundary continuity is set according to the morphological characteristics of ecological boundaries in the ecological prior knowledge base, and a natural continuity constraint requirement is set for the coordinate offset of adjacent boundary points. Then, the semantic context information of the initial boundary pixels is used as the matching basis for variables, requiring that the boundary points corresponding to the variables must maintain a high degree of consistency with the semantic context information of the surrounding areas. The variables, smoothness constraint term, and semantic context information are organically integrated, so that the function can simultaneously respond to the dual requirements of smoothness constraint and semantic context information, and finally construct the energy function of the target area.

[0115] Using the constructed energy function as the core criterion, a minimum graph cut optimization operation is performed on all boundary points on the initial boundary of the candidate ecological redline patch. The coordinate position of each initial boundary point is incorporated into the constraint system of the energy function for verification, determining whether it simultaneously satisfies the coordinate offset requirement of the smoothness constraint term and the matching requirement of semantic context information. For boundary points that do not meet the dual requirements, the coordinate positions are gradually adjusted and then substituted back into the constraint system of the energy function for verification until the boundary point meets the dual constraint requirements. Throughout the optimization process, the constraint feedback of the energy function is always the guide, continuously screening boundary points that can make the constraint fit of the overall boundary of the candidate ecological redline patch in the optimal state. After completing the verification, adjustment and screening of all initial boundary points, the boundary point that finally satisfies the dual constraints and is in the optimal constraint fit is the lowest energy boundary point of the candidate ecological redline patch.

[0116] Extract the spatial coordinate information of all minimum energy boundary points, and sort and arrange them in an orderly manner according to the actual distribution order of each minimum energy boundary point in the geographic space of the target area. In the arrangement process, follow the natural spatial distribution law of the ecological red line boundary in the ecological prior knowledge base. Then, connect all the minimum energy boundary points in the sorted and arranged order. In the connection process, ensure that the connection trajectory between two adjacent minimum energy boundary points meets the requirements of the smoothness constraint, without abrupt coordinate jumps, and the trajectory formed by the entire connection process has no breaks or spatial intersections. Finally, the closed spatial outline formed by connecting all the minimum energy boundary points is the ecological red line boundary of the target area.

[0117] The beneficial effects are as follows: accurate classification of land features is achieved by extracting multi-scale spectral and texture features, generating preliminary boundaries. Based on this, region merging and boundary optimization are performed by combining semantic and spatial relationships to form continuous and accurate candidate ecological redline patches. By constructing an energy function that integrates semantic context and morphological smoothness, and using a graph cut optimization algorithm, the optimal boundary points are selected to ensure the redline's rationality in both semantics and morphology. Finally, after rigorous topological consistency verification and coordinate transformation, an ecological redline vector map conforming to geospatial data specifications is output, significantly improving the scientific rigor and accuracy of redline delineation.

[0118] S4. In the target area, the continuous ecological red line area vector maps are superimposed and compared to identify the spatial distribution information of the ecological red line area vector maps, so as to obtain the set of change area patches in the target area.

[0119] In this embodiment of the invention, the step of overlaying and comparing continuous ecological redline area vector maps in the target area to identify the spatial distribution information of the ecological redline area vector maps, in order to obtain a set of change area patches in the target area, includes:

[0120] The vector maps of the ecological red line areas in two adjacent time phases are superimposed and projected in the same spatial coordinate system to obtain a vector superimposed layer of the target area;

[0121] Spatial topological analysis is performed on the vector overlay layer to obtain a set of candidate change regions for the vector overlay layer.

[0122] Based on the preset change area threshold and change shape rules, the candidate change region set is filtered and eliminated to obtain the effective change region set of the target region;

[0123] The effective change areas are concentrated into independent regions, and the spatial location and boundaries of the independent regions are encapsulated as independent geographic features to obtain the change area patch set of the target area.

[0124] The vector maps of ecological redline areas from two adjacent time periods are uniformly converted to the geographic coordinate system defined for the target area. During the conversion process, the spatial position ratio, boundary outline shape, and coordinate correspondence of the geographic vector elements in each ecological redline area vector map remain unchanged. Then, the vector maps of ecological redline areas from the two time periods that have completed the coordinate system conversion are spatially overlaid according to the standard that the geographic coordinate origin and coordinate scale are completely coincident, so that the geographic vector elements of the ecological redline in the two time periods are fully presented under the same spatial reference system, and finally a vector overlay layer of the target area containing the vector information of the ecological redline areas in the two time periods is formed.

[0125] A point-by-point spatial topological relationship comparison is performed on the geographic vector elements of the ecological red line area in two time phases of the vector overlay layer. The spatial overlap of the area range, outline, and feature coordinate points of each geographic vector element is checked in turn. Areas where the vector elements of the two time phases completely overlap at the area, outline, and coordinate point levels are determined to be unchanged areas. Vector area areas that appear only in one time phase, areas where the outlines of the vector elements of the two time phases have spatial positional offsets, and areas where the spatial range of the vector area increases or decreases are all determined to be changed areas. The spatial range of all the determined changed areas is accurately delineated, and the geographic coordinate range, complete boundary outline, and change type information of each changed area are recorded simultaneously. The set of all the changed areas that have been delineated and recorded is the candidate changed area set of the vector overlay layer.

[0126] The preset change area threshold is the area value of the minimum control patch of the ecological red line determined according to the technical specifications for delineating the ecological red line of the target area. The preset change shape rule is that the outline boundary of the change area is a continuous closed curve, the area outline has no pixel-level fine discrete burrs, and the spatial shape has no isolated point protrusions. For each change area in the candidate change area set, the actual area is calculated and the outline shape is verified in turn. First, the calculated actual area is compared with the preset change area threshold, and the change area with the actual area less than the preset change area threshold is eliminated. Then, for the remaining change areas that meet the area comparison requirements, the outline shape is checked one by one according to the preset change shape rule, and the change area whose outline shape does not meet the change shape rule is eliminated. The change areas that meet the requirements for both area comparison and shape verification are uniformly collected and organized, and the resulting set is the effective change area set of the target area.

[0127] Spatial connectivity is determined for all change regions in the effective change region set, based on the spatial adjacency of geographic coordinates. If two change regions are connected by a continuous geographic outline or partially overlap in spatial area, they are considered connected regions. If two change regions are separated in spatial area and have no geographic outline connection, they are considered independent regions. After determining the spatial connectivity of all change regions, all independent regions are extracted. Geographic element encapsulation is performed on each independent region. During the encapsulation process, the core information such as the geographic spatial coordinate range, characteristic coordinate points of the boundary outline, and the actual measured area value of the independent region are fully preserved. At the same time, each encapsulated geographic element is assigned a unique identification information. All encapsulated independent region geographic elements are systematically integrated to form a set containing all independent geographic elements, which is the change region patch set of the target region.

[0128] The beneficial effects are as follows: by precisely spatially overlaying ecological redline vector maps from two different time periods, a foundational layer for change identification is constructed. Topological analysis based on the overlay results allows for comprehensive and accurate screening of all candidate change areas. Further filtering based on area thresholds and morphological rules eliminates invalid changes, resulting in a set of valid change areas that align with practical management needs. Finally, each independent change area is encapsulated as a uniquely identified geographic element, forming standardized and traceable change patches, providing accurate and reliable data support for the dynamic monitoring of ecological redlines.

[0129] S5. Determine the characteristic change pattern of the patch set in the changed area to obtain the natural evolution state and human activity interference information of the target area;

[0130] In this embodiment of the invention, determining the characteristic change pattern of the patch set in the changed region to obtain the natural evolution state and human activity interference information of the target region includes:

[0131] Key element analysis was performed on the individual changed patches in the changed area to obtain the dominant ecological element type, spatial geometry and temporal span characteristics of the changes in the individual changed patches;

[0132] The dominant ecological element type, spatial geometric morphology, and temporal span characteristics are retrieved and evaluated against a preset natural evolution discrimination rule base and an anthropogenic interference discrimination rule base to obtain the driving type of the changed patch.

[0133] When the driving type is the natural evolution mode, the natural evolution state of the target area is obtained by synchronously analyzing the conversion direction and spatial aggregation degree of the dominant ecological element type.

[0134] When the driving type is the human interference mode, the human activity interference information of the target area is obtained by coordinating the transformation mode of the dominant ecological element type and the spatial distribution intensity.

[0135] For each individual changed patch within the change area cluster, its attribute information, spatial coordinates, and corresponding temporal data are retrieved. The dominant ecological element type is directly obtained from the patch's attribute table; this attribute was assigned during the initial land cover classification. Spatial geometry is obtained by calculating the patch's outline coordinates. Specifically, this involves calculating the ratio of the major axis to the minor axis of the patch's circumscribed rectangle, and simultaneously calculating the ratio of the total length of the patch's boundary line to the square root of its area. These two ratios together describe whether the patch's shape tends towards elongated or regular shapes, and whether its boundaries are complex or smooth. The temporal span of the change is determined by extracting the acquisition dates from the remote sensing images of the two preceding and following time periods of the patch and calculating the difference in the number of days between the two dates.

[0136] The dominant ecological element type, aspect ratio and boundary complexity value of the spatial geometry, and the number of days in time span obtained from the analysis are compared one by one with the preset rule base. The natural evolution discrimination rule base defines multiple patterns. For example, if the dominant ecological element type changes from "herbaceous marshland" to "shrubland" over a time span greater than five years, and the morphology tends towards a naturally expanding irregular shape, then it conforms to the "natural vegetation succession" rule. The anthropogenic interference discrimination rule base also defines multiple patterns. For example, if the dominant ecological element type changes from "woodland" to "construction land" over a time span of less than two years, and the morphology is a regular rectangle or adjacent to road elements, then it conforms to the "construction land expansion" rule. The patch features are matched with all rules to find the rule that satisfies all conditions; the driving type corresponding to this rule is the driving type of the changed patch.

[0137] When the driving type is determined to be a natural evolution pattern, the direction of transformation of the dominant ecological element type is analyzed. The direction of transformation refers to a specific type change such as "water body → mudflat" or "sparse grassland → dense grassland". Spatial aggregation is assessed by calculating the average shortest distance between all patches determined to be naturally evolved within the target area. If this average distance is less than a preset aggregation threshold, it is considered a high-aggregation state; otherwise, it is considered a dispersed state. Ultimately, the natural evolution state of the target area is described by the dominant transformation direction and the degree of aggregation of patches under that transformation direction.

[0138] When the driving type is determined to be an anthropogenic disturbance mode, the dominant ecological element type transformation mode is assessed. Transformation modes refer to type transformations with a clear anthropogenic activity orientation, such as "forest land → industrial and mining land" or "arable land → rural residential land." Spatial distribution intensity is calculated by statistically analyzing the proportion of the total area of ​​patches belonging to the same anthropogenic disturbance mode within the target area to the total area of ​​that area; this proportion is the distribution intensity value. Ultimately, the anthropogenic disturbance information of the target area is jointly described by the specific transformation mode and its corresponding spatial distribution intensity value.

[0139] The beneficial effects are that by analyzing the spatial geometric morphology and temporal span characteristics of the dominant ecological element types of independent changing patches, and accurately determining the driving type based on a preset rule base, reliable information on the natural evolution state and human activity interference of the target area can be obtained, providing a clear basis for the dynamic supervision and scientific decision-making of ecological red lines.

[0140] S6. Integrate the spatial distribution information, the natural evolution status, and the human activity interference information into a dynamic monitoring report of the ecological red line of the target area.

[0141] In this embodiment of the invention, integrating the spatial distribution information, the natural evolution state, and the information on human-induced disturbances into a dynamic monitoring report of the ecological red line for the target area includes:

[0142] By combining the spatial distribution information of the changed patch set, the natural evolution status information, and the human activity interference information, a multi-source dynamic monitoring dataset of the target area is obtained.

[0143] The data in the multi-source dynamic monitoring dataset are unified in terms of spatiotemporal reference, and spatial correlation analysis is performed on the unified dataset to obtain a structured spatiotemporal database of the target area.

[0144] Based on the structured spatiotemporal database and the preset report template framework, the spatial distribution information and the human activity interference information are spatially overlaid and visualized.

[0145] By associating and evaluating the comprehensive information after spatial overlay and visualization rendering with the text describing the natural evolution state, a comprehensive map and a draft of the analysis text of the monitoring results of the target area are obtained.

[0146] The completeness and logical consistency of the comprehensive map and the initial draft of the analysis text of the monitoring results are verified and integrated to obtain the dynamic monitoring report of the ecological red line of the target area.

[0147] The spatial distribution information of each patch in the change area patch set is obtained by collecting its geographic coordinates, area, and unique identifier. The natural evolution status information corresponding to each patch is then read, including a description of the specific evolution direction and aggregation degree. The anthropogenic disturbance information corresponding to each patch is also read, including specific disturbance patterns and distribution intensity values. These three types of information are then grouped according to the patch's unique identifier, with each identifier containing three attributes: spatial, natural evolution, and anthropogenic disturbance. All the grouped data together constitute a multi-source dynamic monitoring dataset for the target area.

[0148] Spatiotemporal benchmark unification was performed on the multi-source dynamic monitoring dataset. Spatial benchmark unification refers to converting all geographic coordinates within the dataset to the "CGCS2000 National Geodetic Coordinate System" to ensure that all spatial information is within the same coordinate framework. Temporal benchmark unification refers to uniformly labeling the time phases corresponding to all information in the format of "monitoring period start date to end date". Spatial correlation analysis was performed on the unified dataset. Specifically, the spatial geometric boundaries, natural evolution status attributes, and anthropogenic disturbance attributes of patches with the same unique identifier were connected and integrated into a database table with fixed fields. This table constitutes the structured spatiotemporal database of the target area, where each record represents a complete changed patch entity.

[0149] The report template framework is a predefined document that specifies the map types, layout, and legend styles that must be included in the dynamic monitoring report of ecological red lines. Based on a structured spatiotemporal database and this template framework, the spatial distribution information of all patches in the database is spatially overlaid with information on human-induced disturbances, and then visualized. Spatial overlay refers to drawing the geometric boundaries of all patches on the same base map. Visualization involves assigning a preset color to each disturbance mode based on the "disturbance mode" field in the human-induced disturbance information; for example, "construction land expansion" is filled with red, and "farmland encroachment" is filled with orange. Thematic maps are then generated according to the legend styles specified in the template.

[0150] The thematic map obtained in the previous step is then correlated and evaluated with the descriptive text regarding the state of natural evolution. The descriptive text is the textual conclusion generated in the previous step regarding the dominant direction and degree of clustering in natural evolution. The specific method for correlation evaluation is to highlight patches belonging to natural evolution on the thematic map with a specific border style, and to clearly indicate the evolutionary state and textual conclusions of these areas in the map description, ensuring a complete correspondence between the map representation and the textual conclusions in spatial location and semantic description. Through this process, a draft of the monitoring results, consisting of a comprehensive visualization map and accompanying analytical text, is obtained.

[0151] The initial draft of the comprehensive monitoring map and analysis text was subjected to completeness and logical consistency checks. Completeness checks involved verifying that the initial draft contained all required sections, charts, and core data items according to the report template framework. Logical consistency checks verified whether the spatial extent of naturally evolved patches represented in the map completely matched the dominant evolutionary direction and specific patch list described in the analysis text, and also verified whether the statistical area of ​​human-caused interference patches was consistent with the numerical calculation results of the distribution intensity given in the text. All missing items identified during the checks were supplemented, and all inconsistencies were corrected according to the source data in the structured spatiotemporal database. The corrected maps and text were then merged and arranged to generate a complete and internally consistent dynamic monitoring report of the ecological red line for the target area.

[0152] The beneficial effects are that by systematically collecting multi-source monitoring data and unifying their spatiotemporal benchmarks, a structured spatiotemporal database is constructed. Then, based on standardized templates, a draft of the monitoring results is generated that integrates spatial visualization information and text analysis conclusions. Finally, after rigorous verification to ensure the completeness of the content and the consistency of logic, a reliable and standardized dynamic monitoring report of the ecological red line is generated.

[0153] like Figure 2 The diagram shown is a functional block diagram of an ecological red line dynamic monitoring system based on remote sensing images provided in an embodiment of the present invention.

[0154] The ecological red line dynamic monitoring system 100 based on remote sensing images described in this invention can be installed in an electronic device. Depending on the functions implemented, the ecological red line dynamic monitoring system 100 based on remote sensing images may include a data preprocessing and input module 101, a feature extraction module 102, a change initial detection module 103, a change area vectorization module 104, a change pattern determination module 105, and a result integration and output module 106. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0155] In this embodiment, the functions of each module / unit are as follows:

[0156] The data preprocessing and input module 101 is used to preprocess the multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area.

[0157] The feature extraction module 102 is used to extract ecological element characteristics from the standard image sequence to obtain the spectral features and spatial texture features of the target region.

[0158] The change detection module 103 is used to perform edge optimization segmentation on the standard image sequence based on the spectral features and spatial texture features to obtain the ecological red line region vector map of the target region;

[0159] The change area vectorization module 104 is used to overlay and compare the continuous ecological red line area vector maps in the target area to identify the spatial distribution information of the ecological red line area vector maps, so as to obtain the change area patch set of the target area.

[0160] The change pattern determination module 105 is used to determine the feature change pattern of the change region patch set to obtain the natural evolution state and human activity interference information of the target region.

[0161] The result integration and output module 106 is used to integrate the spatial distribution information, the natural evolution state, and the human activity interference information into a dynamic monitoring report of the ecological red line of the target area.

[0162] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0163] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0164] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0165] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0166] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for dynamic monitoring of ecological red lines based on remote sensing images, characterized in that, The method includes: S1. Perform data preprocessing on the multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area; S2. Extract ecological element characteristics from the standard image sequence to obtain the spectral and spatial texture features of the target region; S3. Based on the spectral features and spatial texture features, perform edge optimization segmentation on the standard image sequence to obtain a vector map of the ecological redline region of the target area, including: The pixel-level semantic features in the standard image sequence are subjected to full-image scanning classification to obtain a preliminary land cover classification map of the target area. Based on the pixel row and column arrangement order of the standard image sequence, starting from the first pixel in the upper left corner of the image, the entire image is scanned row by row and column by column until the last pixel in the lower right corner of the image. During the scanning process, the corresponding spectral features and spatial texture features are extracted for each pixel to form the pixel-level semantic features of that pixel. Edge detection is performed on the preliminary land feature classification map to obtain the preliminary boundary pixels of the target area; Based on the semantic category and spatial proximity of the preliminary boundary pixels, the preliminary land cover classification map is subjected to region growing and merging to obtain candidate ecological redline patches for the target area; The candidate ecological redline patches are finely optimized to obtain the ecological redline boundary of the target area; Based on the topological consistency of the target area, the ecological red line boundary is transformed into a geographic vector map to obtain the ecological red line area vector map of the target area. The step of finely optimizing the candidate ecological redline patches to obtain the ecological redline boundary of the target area includes: Multi-scale deep feature extraction is performed on the surrounding area of ​​the candidate ecological red line patch to obtain the semantic context information of the preliminary boundary pixels; Based on a pre-defined ecological prior knowledge base, the energy function of the target region is constructed using the coordinates of the boundary points of the initial boundary in the candidate ecological red line patch as variables, and the smoothness constraint term of the boundary continuity in the ecological prior knowledge base and the semantic context information as common constraints. The candidate ecological redline patches are optimized by minimizing graph cut using the energy function to obtain the lowest energy boundary point of the candidate ecological redline patches. The ecological red line boundary of the target area is determined based on the minimum energy boundary point. S4. In the target area, the continuous ecological red line area vector maps are superimposed and compared to identify the spatial distribution information of the ecological red line area vector maps, so as to obtain the set of change area patches in the target area. S5. Determine the characteristic change pattern of the patch set in the changed area to obtain the natural evolution state and human activity interference information of the target area; S6. Integrate the spatial distribution information, the natural evolution status, and the human activity interference information into a dynamic monitoring report of the ecological red line of the target area.

2. The method for dynamic monitoring of ecological red lines based on remote sensing images as described in claim 1, characterized in that, The process of preprocessing multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area includes: Radiometric calibration is performed on multi-temporal remote sensing image data of the target area to obtain radiometric image data of the multi-temporal remote sensing image data; Atmospheric correction is performed on the radiance image data to obtain the surface reflectance image of the target area; Based on the preset digital elevation model and ground control points in the target area, the geometric deformation in the surface reflectance image is corrected to obtain the orthorectified image data of the target area. The orthorectified images are registered and cropped to align and unify the orthorectified images from different time phases to the boundary of the target region, thus obtaining a standard image sequence of the target region.

3. The method for dynamic monitoring of ecological red lines based on remote sensing images as described in claim 1, characterized in that, The step of extracting ecological element characteristics from the standard image sequence to obtain the spectral and spatial texture features of the target region includes: Super-resolution reconstruction is performed on the standard image sequence to obtain enhanced image data of the target region; Based on the multi-band reflectance data in the standard image sequence, statistical analysis is performed on the land cover type index of the target area to obtain the spectral characteristics of the target area. The enhanced image data is analyzed for grayscale spatial relationship features to obtain the spatial texture features of the target region.

4. The method for dynamic monitoring of ecological red lines based on remote sensing images as described in claim 1, characterized in that, The formulas for calculating the relevant energy boundary points in the energy function are as follows: ； In the formula, The set of candidate boundary points in the candidate ecological redline patches The overall energy value, The preset semantic balance weight coefficients, Boundary points in the initial boundary The degree of consistency of semantic context information The preset adaptive smoothness weight coefficients, For semantic context confidence, For the smoothness constraint term of the ecological red line boundary, Boundary point The set of neighboring pixels; The degree of agreement of the semantic context information of the boundary point comes from the quantitative result of matching the semantic context information extracted from the multi-scale deep features corresponding to the boundary point with the feature information of the ecological red line land cover category in the preset ecological prior knowledge base. During the matching process, the overlap of feature information is used as the basis to numerically represent the agreement between the semantic information of the boundary point and the prior features. Semantic context confidence is derived from a quantitative assessment of the effectiveness and reliability of the semantic context information corresponding to the boundary point. The assessment basis is the completeness of multi-scale deep feature extraction in the area surrounding the boundary point. The more complete the extracted feature information dimensions, the higher the confidence value. The assessment basis is formulated based on the technical specifications of multi-scale feature extraction to ensure that the assessment results can truly reflect the quality of semantic context information.

5. The method for dynamic monitoring of ecological red lines based on remote sensing images as described in claim 1, characterized in that, In the target area, continuous ecological redline area vector maps are overlaid and compared to identify the spatial distribution information of the ecological redline area vector maps, so as to obtain a set of change area patches in the target area, including: The vector maps of the ecological red line areas in two adjacent time phases are superimposed and projected in the same spatial coordinate system to obtain a vector superimposed layer of the target area; Spatial topological analysis is performed on the vector overlay layer to obtain a set of candidate change regions for the vector overlay layer. Based on the preset change area threshold and change shape rules, the candidate change region set is filtered and eliminated to obtain the effective change region set of the target region; Extract independent regions from the effective change region set, and encapsulate the spatial location and boundaries of the independent regions as independent geographic features to obtain the change region patch set of the target region.

6. The method for dynamic monitoring of ecological red lines based on remote sensing images as described in claim 1, characterized in that, The step of determining the characteristic change pattern of the patch set in the changed region to obtain the natural evolution state and human activity interference information of the target region includes: Key element analysis was performed on the isolated changed patches in the changed area to obtain the dominant ecological element type, spatial geometry and time span characteristics of the changes in the isolated changed patches; The dominant ecological element type, spatial geometric morphology, and temporal span characteristics are retrieved and evaluated against a preset natural evolution discrimination rule base and an anthropogenic interference discrimination rule base to obtain the driving type of the changed patch. When the driving type is the natural evolution mode, the natural evolution state of the target area is obtained by synchronously analyzing the conversion direction and spatial aggregation degree of the dominant ecological element type. When the driving type is the human interference mode, the human activity interference information of the target area is obtained by coordinating the transformation mode of the dominant ecological element type and the spatial distribution intensity.

7. The method for dynamic monitoring of ecological red lines based on remote sensing images as described in claim 1, characterized in that, The process of integrating the spatial distribution information, the natural evolution status, and the information on human-induced disturbances into a dynamic monitoring report of the ecological red line for the target area includes: By combining the spatial distribution information of the changed patch set, the natural evolution status information, and the human activity interference information, a multi-source dynamic monitoring dataset of the target area is obtained. The data in the multi-source dynamic monitoring dataset are unified in terms of spatiotemporal reference, and spatial correlation analysis is performed on the unified dataset to obtain a structured spatiotemporal database of the target area. Based on the structured spatiotemporal database and the preset report template framework, the spatial distribution information and the human activity interference information are spatially overlaid and visualized. The integrated information, which has undergone spatial overlay and visualization rendering, is correlated and evaluated with the descriptive text of the natural evolution state to obtain a comprehensive map and a draft of the analysis text of the monitoring results of the target area. The completeness and logical consistency of the comprehensive map and the initial draft of the analysis text of the monitoring results are verified and integrated to obtain the dynamic monitoring report of the ecological red line of the target area.

8. A dynamic monitoring system for ecological red lines based on remote sensing images, characterized in that, The system for implementing the dynamic monitoring method for ecological red lines based on remote sensing images as described in claim 1 includes: The data preprocessing and input module is used to preprocess multi-temporal remote sensing image data of the target area to obtain a standard image sequence of the target area. The feature extraction module is used to extract ecological element characteristics from the standard image sequence to obtain the spectral features and spatial texture features of the target region; The change detection module is used to perform edge optimization segmentation on the standard image sequence based on the spectral features and spatial texture features to obtain a vector map of the ecological redline region of the target area, including: The pixel-level semantic features in the standard image sequence are subjected to full-image scanning classification to obtain a preliminary land cover classification map of the target area. Based on the pixel row and column arrangement order of the standard image sequence, starting from the first pixel in the upper left corner of the image, the entire image is scanned row by row and column by column until the last pixel in the lower right corner of the image. During the scanning process, the corresponding spectral features and spatial texture features are extracted for each pixel to form the pixel-level semantic features of that pixel. Edge detection is performed on the preliminary land feature classification map to obtain the preliminary boundary pixels of the target area; Based on the semantic category and spatial proximity of the preliminary boundary pixels, the preliminary land cover classification map is subjected to region growing and merging to obtain candidate ecological redline patches for the target area; The candidate ecological redline patches are finely optimized to obtain the ecological redline boundary of the target area; Based on the topological consistency of the target area, the ecological red line boundary is transformed into a geographic vector map to obtain the ecological red line area vector map of the target area. The step of finely optimizing the candidate ecological redline patches to obtain the ecological redline boundary of the target area includes: Multi-scale deep feature extraction is performed on the surrounding area of ​​the candidate ecological red line patch to obtain the semantic context information of the preliminary boundary pixels; Based on a pre-defined ecological prior knowledge base, the energy function of the target region is constructed using the coordinates of the boundary points of the initial boundary in the candidate ecological red line patch as variables, and the smoothness constraint term of the boundary continuity in the ecological prior knowledge base and the semantic context information as common constraints. The candidate ecological redline patches are optimized by minimizing graph cut using the energy function to obtain the lowest energy boundary point of the candidate ecological redline patches. The ecological red line boundary of the target area is determined based on the minimum energy boundary point. The change area vectorization module is used to overlay and compare continuous ecological red line area vector maps in the target area to identify the spatial distribution information of the ecological red line area vector maps, so as to obtain the change area patch set of the target area; The change pattern determination module is used to determine the feature change patterns of the patch set in the changed region to obtain the natural evolution state and human activity interference information of the target region. The results integration and output module is used to integrate the spatial distribution information, the natural evolution status, and the human activity interference information into a dynamic monitoring report of the ecological red line of the target area.

Images