Cross-grid patch adaptive merging method and system for grid-based remote sensing interpretation

Abstract

The application provides a cross-grid image spot adaptive merging method and system for gridding remote sensing interpretation, and belongs to the technical field of computer vision. The method first performs external expansion buffer preprocessing on a basic interpretation grid to generate an expanded interpretation unit with an overlapping area. Then, the ratio of the intersection area of spatially adjacent image spots to the minimum area is calculated as the overlap degree, and the merging relationship is automatically determined by comparing with a preset threshold. A merging relationship graph is constructed based on the determination result, and all image spot sets to be merged are identified by using a connected component searching algorithm. For each set, the vertices thereof are merged and the contour is reconstructed to generate a polygon representing the overall geometric shape of the ground object. Finally, the ground object category is inherited through a confidence weighted voting mechanism, and the attribute information such as the confidence is fused by using an extreme value strategy. The application can accurately merge fragmented image spots caused by grid division or recognition errors, and significantly improves the geometric integrity and accuracy of remote sensing interpretation results.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer vision, and in particular to a cross-grid patch adaptive merging method and system for grid-based remote sensing interpretation. BACKGROUND

[0002] With the wide application of satellite remote sensing technology, artificial intelligence (AI) technology is widely applied to remote sensing interpretation. However, due to the scope of single AI interpretation, the grid-based (block) processing strategy is adopted for automatic interpretation of large-scale remote sensing images, which has become the mainstream method in the industry. However, this strategy has an inherent defect: the ground objects (such as bridges and highways) are often cut by grid boundaries, forming multiple fragmented interpretation patches in adjacent grids. This leads to problems including:

[0003] (1) Incomplete ground object information: subsequent analysis cannot accurately obtain the overall properties (such as type, total number, and total area) of the ground object;

[0004] (2) Reduced recognition accuracy: the confidence of fragmented patches is usually lower than that of complete patches, affecting the accuracy of AI judgment;

[0005] (3) Low analysis efficiency: manual post-processing is required to merge the fragments, and full automation cannot be achieved.

[0006] In the prior art, although there are general vector face merging tools (such as the Dissolve and Merge functions in GIS software), they lack intelligence and can only be based on simple attributes (such as ground object types) for merging, and cannot determine whether multiple patches from different grids belong to the same physical ground object, which easily leads to false merging or missing merging.

[0007] Therefore, there is an urgent need for a special algorithm that can automatically, accurately, and efficiently identify and merge cross-grid fragmented patches from the same ground object.

[0008] In addition, the existing grid-based interpretation strategy usually strictly cuts along the grid boundaries without considering the impact of ground object integrity on subsequent analysis. This makes the probability of cutting off ground objects located at the edge of the grid very high, and the fragmented patches often lack sufficient overlapping areas, resulting in the failure of the subsequent automatic merging module due to the lack of reliable spatial topological relationships, and forced to rely on manual intervention. SUMMARY

[0009] The present application provides a cross-grid patch adaptive merging method and system for grid-based remote sensing interpretation to solve at least one defect in the prior art.

[0010] In a first aspect, the present application provides a cross-grid polygon adaptive merging method for grid-based remote sensing interpretation, comprising: performing an interpretation range expansion preprocessing on the interpretation grid of a remote sensing image; based on the interpretation grid after the expansion preprocessing, obtaining a plurality of polygons to be merged, each polygon comprising geometric polygon data and attribute information, the attribute information comprising a feature category and a confidence level; for two spatially adjacent polygons, calculating the spatial intersection area thereof, calculating an overlap degree based on the minimum value of the spatial intersection area and the area of the two polygons, comparing the overlap degree with a preset threshold, and if the overlap degree is greater than or equal to the preset threshold, determining that the two polygons should be merged; based on the determination results of all the polygons to be merged, constructing an undirected graph taking the polygons as nodes and the merging relationship as edges as a merging relationship graph, traversing the merging relationship graph, and identifying all connected components, each connected component representing a polygon set to be merged; for each connected component, merging the geometric vertices of all polygons in the connected component into a unified vertex set, and applying a contour reconstruction algorithm to the unified vertex set to generate a merged polygon representing the overall geometric shape; based on the attribute information of the polygons in the connected component, determining the feature category of the merged polygon through a voting mechanism, and taking the maximum value of the confidence levels of the polygons in the connected component as the confidence level of the merged polygon; and outputting the merged polygon.

[0011] According to the cross-grid polygon adaptive merging method for grid-based remote sensing interpretation provided by the present application, the interpretation range expansion preprocessing is performed on the interpretation grid of a remote sensing image, comprising: performing an expansion buffer on the interpretation range of each basic grid to generate a larger rectangular interpretation unit concentric with the basic grid as the interpretation grid after the preprocessing; wherein the expansion width is a fixed proportion of the side length of the basic grid.

[0012] According to the cross-grid polygon adaptive merging method for grid-based remote sensing interpretation provided by the present application, the method for calculating the overlap degree comprises: dividing the spatial intersection area of the two polygons by the minimum value of the areas of the two polygons.

[0013] According to the cross-grid polygon adaptive merging method for grid-based remote sensing interpretation provided by the present application, the contour reconstruction algorithm adopts a convex hull algorithm or a concave hull algorithm.

[0014] According to the cross-grid polygon adaptive merging method for grid-based remote sensing interpretation provided by the present application, the voting mechanism is a confidence level weighted voting mechanism, specifically, a weighted score is calculated according to the feature category label, and the category with the highest score is taken as the category of the merged polygon.

[0015] According to the cross-grid polygon adaptive merging method for grid-based remote sensing interpretation provided by the present application, the preset threshold is dynamically configured based on different feature types.

[0016] The self-adaptive merging method for cross-grid image patches for grid-based remote sensing interpretation provided by the application is based on an Alpha Shape algorithm as the concave hull algorithm and a Convex Hull algorithm as the convex hull algorithm.

[0017] In a second aspect, the application further provides a self-adaptive merging system for cross-grid image patches for grid-based remote sensing interpretation, comprising:

[0018] The first processing module is configured to perform an expansion preprocessing of the interpretation range of the interpretation grid of the remote sensing image.

[0019] The second processing module is configured to obtain a plurality of image patches to be merged based on the interpretation grid after the expansion preprocessing, each image patch comprising geometric polygon data and attribute information, and the attribute information comprising a feature category and a confidence level.

[0020] The third processing module is configured to calculate the spatial intersection area of two spatially adjacent image patches, calculate the overlap degree based on the minimum value of the spatial intersection area and the area of the two image patches, compare the overlap degree with a preset threshold, and determine that the two image patches should be merged if the overlap degree is greater than or equal to the preset threshold.

[0021] The fourth processing module is configured to construct a merge relationship graph by taking the image patches as nodes and the merge relationship as edges based on the determination results of all the image patches that should be merged, traverse the merge relationship graph, and identify all the connected components, each connected component representing a set of image patches to be merged.

[0022] The fifth processing module is configured to merge the geometric vertices of all the image patches in each connected component into a unified vertex set, and apply a contour reconstruction algorithm to the unified vertex set to generate a merged polygon representing the overall geometric shape.

[0023] The sixth processing module is configured to determine the feature category of the merged image patch based on the attribute information of the image patches in the connected component through a voting mechanism, and take the maximum value of the confidence levels of the image patches in the connected component as the confidence level of the merged image patch.

[0024] The seventh processing module is configured to output the merged image patch.

[0025] In a third aspect, the application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the self-adaptive merging method for cross-grid image patches for grid-based remote sensing interpretation as described above.

[0026] In a fourth aspect, the present application also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method for adaptive merging of cross-grid map patches for grid-based remote sensing interpretation.

[0027] The method and system for adaptive merging of cross-grid map patches for grid-based remote sensing interpretation provided by the present application have the following advantages over the prior art:

[0028] (1) Automation and high precision: By quantifying the overlap index and connected component analysis, the merging relationship can be accurately determined without human intervention, significantly reducing false merging and missed merging.

[0029] (2) High shape fidelity: The concave / convex hull algorithm is used for contour reconstruction, which can better maintain the original geometric features of the features than simple "fusion".

[0030] (3) Semantic information fusion: The confidence of AI interpretation is innovatively integrated into the attribute inheritance strategy, making the class judgment of the merged patches more reliable.

[0031] (4) Strong universality: The method is not limited to river and lake supervision and soil and water conservation supervision in the water conservancy industry, and can be widely applied to land resource investigation, urban planning, agricultural census and any scenario using grid-based remote sensing interpretation, having good universality and promotion value. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0033] Figure 1 is a flowchart of the method for adaptive merging of cross-grid map patches for grid-based remote sensing interpretation provided by the present application;

[0034] Figure 2 is a schematic diagram of grid expansion provided by the present application;

[0035] Figure 3 is a comparison diagram of interpretation results provided by the present application;

[0036] Figure 4 is a schematic diagram of the structure of an electronic device provided by the present application. DETAILED DESCRIPTION

[0037] In order to make the objects, technical solutions and advantages of the present application clearer, the following will clearly and completely describe the technical solutions in the present application with reference to the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0038] It should be noted that, in the description of the embodiments of the present application, the terms "comprising", "containing" or any other variants thereof are intended to cover the non-exclusive containing, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitation, the element defined by the sentence "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0039] The terms "first", "second", and the like used in the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than that illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a category, not limited to the number of objects, for example, the first object can be one or more. In addition, "and / or" means at least one of the connected objects, and the character " / ", generally means that the front and rear associated objects are in an "or" relationship.

[0040] The following will be described in detail with reference to the drawings. Figures 1-4 The method and system for adaptive cross-grid polygon merging for grid-based remote sensing interpretation provided by the embodiments of the present application are described.

[0041] Figure 1 The flowchart of the method for adaptive cross-grid polygon merging for grid-based remote sensing interpretation provided by the present application is shown in FIG. 1, which includes but is not limited to the following steps: Figure 1

[0042] Step 101: Preprocessing the expansion of the interpretation range of the interpretation grid (basic grid) of the remote sensing image.

[0043] In order to solve the problem of cutting of ground objects at the edge of the grid and create necessary conditions for subsequent merging, the interpretation range is first preprocessed and expanded.

[0044] The edge length of the grid is determined according to the interested range of the remote sensing image (which can be the management range of rivers and lakes), which is 2m.​ Basic grid After (e.g., 1000m x 1000m), the interpretation range of each grid is buffered to construct a larger rectangular interpretation unit concentric with it. That is, the interpretation mesh after preprocessing.

[0045] The extension rule is the interpretation unit. The boundary is formed by the base grid. The boundary is buffered outwards by a fixed width. get. The value of is a key technical parameter, set to the original mesh side length ( A fixed ratio () The calculation formula is as follows:

[0046]

[0047] Among them, the external expansion ratio The range of values ​​is Preferred, .

[0048] Figure 2 A schematic diagram of the mesh expansion provided by this invention is shown below. Figure 2 As shown, the red network is the base grid, and the green network is the outer extension.

[0049] The present invention has the following beneficial effects through external expansion pretreatment:

[0050] (1) Reduce edge cutting: When the main body of the ground feature is located at the edge of the grid, because part of it falls into the outer expansion area, it can still be interpreted relatively completely by the AI ​​model, thus significantly reducing the probability of the patch being cut off by the grid boundary.

[0051] (2) Creating overlapping areas: For a feature that spans two grids, it is distributed within the interpretation range of two adjacent grids. Since each grid is extended outward, the feature can be interpreted as a patch in both grids, and an overlapping zone will be formed within the extended buffer zone of the two adjacent grids. This ensures that the patches interpreted by the feature in the two grids will inevitably intersect, providing a crucial spatial topological basis for subsequent automatic merging.

[0052] Step 102: Based on the interpretation grid after expansion preprocessing, obtain multiple patches to be merged. Each patch includes geometric polygon data and attribute information, including land cover category and confidence level.

[0053] In the expanded interpretation unit Inside, the pre-trained remote sensing image interpretation model is used for automatic identification of ground objects, and the initial interpretation polygon of the target ground object (such as a wharf, a bridge, etc.) is extracted . Each polygon contains geometric polygon information and its corresponding ground object category label and prediction confidence .

[0054] Step 103: For two spatially adjacent polygons, calculate the spatial intersection area, calculate the overlap degree based on the minimum value of the spatial intersection area and the area of the two polygons, compare the overlap degree with a preset threshold, and if the overlap degree is greater than or equal to the preset threshold, determine that the two polygons should be merged.

[0055] This step is the core of determining which polygons should be merged.

[0056] (1) Calculate the intersection of the polygons: for any two polygons and from adjacent grids (extended and preprocessed interpretation grids) , calculate the spatial geometric intersection , and calculate the intersection area .

[0057] (2) Calculate the overlap degree index: define and calculate the overlap degree based on the minimum area:

[0058]

[0059] This index measures the proportion of the smaller polygon covered by the other polygon, which can effectively avoid the large polygon swallowing the small polygon with only a small intersection.

[0060] (3) Merge determination: set an overlap degree threshold (the experience value range , preferably ). If it satisfies:

[0061]

[0062] , then determine that the polygons and come from the same physical ground object and should be merged.

[0063] Step 104: Based on all the determination results of the polygons that should be merged, construct an undirected graph with the polygons as nodes and the merging relationship as edges as the merging relationship graph, traverse the merging relationship graph, and identify all connected components, each connected component representing a set of polygons to be merged.

[0064] Take all interpretation polygons as the vertex set . For any two polygons, if they satisfy the above merging criterion An edge is then established between them Thus a undirected graph is constructed That is, a merged graph.

[0065] Step 105: For each connected component, merge the geometric vertices of all the patches in the connected component into a unified vertex set, and apply a contour reconstruction algorithm to the unified vertex set to generate a merged polygon representing the overall geometric shape.

[0066] Apply a graph theory algorithm (such as Depth-First Search DFS or Union-Find algorithm) to find all the connected components in the graph Each connected component represents a set of patches that need to be merged.

[0067] For each connected component to be merged:

[0068] (1) Vertex set merging: merge the polygon vertices of all the patches in the set into a new vertex set , which is expressed as:

[0069]

[0070] wherein represents the geographic coordinates of a vertex.

[0071] (2) Contour generation: process the vertex set by a geometric contour reconstruction algorithm to generate a merged polygon that encloses all the vertices. Specifically, apply an Alpha Shape algorithm or a Convex Hull algorithm to the vertex set to generate a new polygon that encloses all the vertices as the merged geometric contour.

[0072] Preferably, the Alpha Shape algorithm can better fit the actual concave shape of the ground object and has high restoration degree, and its effect is better than that of simple convex hull fusion. The Convex Hull algorithm or other geometric algorithms that can realize contour reconstruction can also be used. The Convex Hull algorithm has relatively high calculation efficiency compared with the Alpha Shape algorithm, but it can oversimplify the contour.

[0073] The formula of the Convex Hull algorithm is:

[0074]

[0075] ​The concave hull (Alpha Shape) algorithm can better reconstruct the real shape of ground objects that may have concave features. It can be regarded as a generalization of the convex hull, with a scale parameter to control the fineness of the generated contour. Its basic principle is to identify pairs of points in the point set that satisfy the empty circle condition and connect them into edges, thereby generating an outer envelope polygon that may contain concave features. When , the concave hull approaches the convex hull.

[0076] Step 106: Based on the attribute information of each patch in the connected component, determine the ground object class of the merged patch through a voting mechanism, and take the maximum value of the confidence of each patch in the connected component as the confidence of the merged patch.

[0077] To the new merged patch attribute is the key to its effective use.

[0078] Ground object class label inheritance (confidence weighted voting): The class label of the new patch is determined by the label of all patches in the set through confidence weighted voting. The class of the new patch is the one that maximizes the sum of the confidence of all original patches belonging to that class. The calculation formula is:

[0079]

[0080] where,

[0081] is the set of all possible ground object classes (such as {port, bridge,...});

[0082] is the predicted confidence of the patch ;

[0083] is the Kronecker delta, used to determine whether the labels are consistent:

[0084]

[0085] The confidence of the new patch after merging can take the maximum or average value of the confidence of the patches in the original set:

[0086] .

[0087] Step 107: Output the merged patch data, which includes the merged polygon, the inherited ground object class, and the confidence.

[0088] Finally output the vector set of all independent patches and merged patches and their corresponding attributes, completing the entire processing flow.

[0089] Figure 3 is the comparison diagram of the interpretation result provided by the present application, as shown in the figure, the traditional interpretation result is ACIK and CEFI, and the interpretation result of the present application is ADGK and BEFJ, and an intersection (overlapping area) BDGJ is formed at the extended boundary, which can be considered as the same ground object, and the merged patch is AEFK. Figure 3

[0090] The cross-grid patch adaptive merging method for grid-based remote sensing interpretation provided by the present application has the following beneficial effects compared with the prior art:

[0091] (1) Automation and high precision: Through quantitative overlap index and connected component analysis, the merging relationship can be accurately judged without manual intervention, significantly reducing false merging and missing merging.

[0092] (2) High shape fidelity: The concave / convex hull algorithm is used for contour reconstruction, which can better maintain the original geometric features of the ground object than simple "fusion".

[0093] (3) Semantic information fusion: The confidence of AI interpretation is innovatively integrated into the attribute inheritance strategy, making the category judgment of the merged patch more reliable.

[0094] (4) Strong universality: The method is not limited to river and lake supervision and soil and water conservation supervision in the water conservancy industry, and can be widely applied to land and resource investigation, urban planning, agricultural census and any scene using grid-based remote sensing interpretation, and has good universality and promotion value.

[0095] Taking a bridge patch merging across two adjacent grids as an example, the specific implementation process of the patch merging algorithm patent is described in detail. This embodiment only relates to the patch merging algorithm itself, and does not involve the previous grid division, AI interpretation and subsequent compliance judgment steps.

[0096] First step: input the fragmented patches to be merged

[0097] The algorithm receives two bridge fragment patches and their attribute information output from the upstream processing module (such as the AI interpretation module):

[0098] Patch from grid A, {geometric polygon: vertex set , category: bridge, confidence: 0.96, area: 850 }

[0099] Patch ​: from grid B, {geometry polygon: vertex set , category: bridge, confidence: 0.93, area: 780 }

[0100] Step 2: Spatial Topology Analysis (Calculate Merge Criteria)

[0101] The algorithm automatically performs the following calculations:

[0102] Calculate Spatial Intersection: Call the spatial geometry calculation library (such as GEOS) to calculate and Polygon Intersection , and calculate its area .

[0103] Calculate Minimum Area-based Overlap (OR_min):

[0104]

[0105] Merge Decision: Compare the calculation results with the preset threshold (take in this example). Since , it is determined that and come from the same physical bridge and should be merged.

[0106] Step 3: Build a Merge Relationship Graph and Find Connected Components

[0107] Build a Merge Relationship Graph: Take and as nodes and establish an undirected edge between them to form a simple merge relationship graph .

[0108] Find Connected Components: Apply the Union-Find algorithm to find connected components in the graph. The algorithm confirms that and belong to the same connected component .

[0109] Step 4: Geometric Merge and Contour Reconstruction

[0110] Geometric Merge: Merge all the vertices of the connected component into a new vertex set .

[0111]

[0112] Contour Reconstruction: Reconstruct the vertex set The concave hull (Alpha Shape) algorithm (parameters ) is used for contour reconstruction to generate a new single polygon that can wrap all input vertices . The polygon accurately restores the overall geometry of the bridge, and its calculated area is .

[0113] Step 5: Attribute and semantic information inheritance

[0114] Feature class label inheritance: Since the class labels of and are both "bridge", a confidence weighted voting mechanism is used:

[0115]

[0116] Confidence fusion: the maximum value strategy is used to inherit the highest confidence in the original polygon to represent the confidence degree of the merged result:

[0117]

[0118] Step 6: Output

[0119] The algorithm ends and outputs the final polygon after merging:

[0120] {

[0121] Geometric polygon: ` (vertex sequence)`,

[0122] Class: `bridge`,

[0123] Confidence: `0.96`,

[0124] Area: `1630 `

[0125] }

[0126] The result can be output to a downstream storage system or visualization module to complete the merging operation.

[0127] On the other hand, the present application also provides a cross-grid polygon adaptive merging system for gridded remote sensing interpretation, comprising:

[0128] A first processing module is configured to perform an interpretation range expansion preprocessing on the interpretation grid of the remote sensing image;

[0129] A second processing module is configured to obtain a plurality of polygons to be merged based on the interpretation grid after the expansion preprocessing, each polygon comprising geometric polygon data and attribute information, and the attribute information comprising feature class and confidence;

[0130] The third processing module is configured to calculate a spatial intersection area of two spatially adjacent map patches, calculate an overlap degree based on a minimum value of the spatial intersection area and areas of the two map patches, compare the overlap degree with a preset threshold, and determine that the two map patches should be merged if the overlap degree is greater than or equal to the preset threshold.

[0131] The fourth processing module is configured to construct a merge relationship graph by taking the map patches as nodes and the merge relationship as edges based on all determination results of the map patches that should be merged, traverse the merge relationship graph, and identify all connected components, each of which represents a set of map patches to be merged.

[0132] The fifth processing module is configured to, for each connected component, merge geometric vertices of all map patches in the connected component into a unified vertex set, and apply a contour reconstruction algorithm to the unified vertex set to generate a merged polygon representing an overall geometric shape.

[0133] The sixth processing module is configured to determine a feature class of the merged map patch based on attribute information of the map patches in the connected component through a voting mechanism, and take a maximum value of confidences of the map patches in the connected component as a confidence of the merged map patch.

[0134] The seventh processing module is configured to output the merged map patch.

[0135] It should be noted that the cross-grid map patch adaptive merging system for grid-based remote sensing interpretation provided in the embodiments of the present application can execute the cross-grid map patch adaptive merging method for grid-based remote sensing interpretation described in any of the above embodiments when actually running, and thus the embodiments of the present application will not be described herein.

[0136] Figure 4 is a structural schematic diagram of an electronic device provided by the present application, as shown in Figure 4 The electronic device can include a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 can communicate with each other through the communications bus 440. The processor 410 can invoke a logical instruction in the memory 430 to execute the cross-grid map patch adaptive merging method for grid-based remote sensing interpretation.

[0137] In addition, the logic instructions in the memory 430 described above can be implemented in the form of software function units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0138] In another aspect, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer can execute the cross-grid plot adaptive merging method for grid-based remote sensing interpretation provided by the above-mentioned embodiments.

[0139] In another aspect, the present application also provides a non-transitory computer readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the cross-grid plot adaptive merging method for grid-based remote sensing interpretation provided by the above-mentioned embodiments is implemented.

[0140] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A cross-grid patch adaptive merging method for gridded remote sensing interpretation, characterized in that, include: The interpretation grid of the remote sensing image is preprocessed to expand the interpretation range. Based on the interpretation grid after expansion preprocessing, multiple patches to be merged are obtained. Each patch includes geometric polygon data and attribute information, including land cover category and confidence level. For two spatially adjacent patches, calculate their spatial intersection area, calculate the overlap degree based on the minimum value of the spatial intersection area and the area of ​​the two patches, compare the overlap degree with a preset threshold, and if the overlap degree is greater than or equal to the preset threshold, determine that the two patches should be merged. Based on all the determination results of merging, an undirected graph with patches as nodes and merging relationships as edges is constructed as the merging relationship graph. The merging relationship graph is traversed to identify all connected components, and each connected component represents a set of patches to be merged. For each connected component, the geometric vertices of all features within that connected component are merged into a unified vertex set. A contour reconstruction algorithm is then applied to the unified vertex set to generate a merged polygon representing the overall geometry. Based on the attribute information of each patch within the connected component, the land feature category of the merged patch is determined through a voting mechanism, and the maximum confidence score of each patch within the connected component is taken as the confidence score of the merged patch. Output the merged patch.

2. The cross-grid patch adaptive merging method for gridded remote sensing interpretation according to claim 1, characterized in that, The interpretation grid of the remote sensing image undergoes an expansion preprocessing step to extend the interpretation range, including: The interpretation range of each basic grid is buffered outward to generate a larger rectangular interpretation cell concentric with it, which serves as the preprocessed interpretation grid. The outward expansion width is a fixed proportion of the base grid side length.

3. The cross-grid patch adaptive merging method for gridded remote sensing interpretation according to claim 1, characterized in that, Methods for calculating overlap include: Divide the area of ​​the spatial intersection of two patches by the minimum area of ​​the two patches.

4. The cross-grid patch adaptive merging method for gridded remote sensing interpretation according to claim 1, characterized in that, The contour reconstruction algorithm employs either the concave hull algorithm or the convex hull algorithm.

5. The cross-grid patch adaptive merging method for gridded remote sensing interpretation according to claim 1, characterized in that, The voting mechanism is a confidence-weighted voting mechanism, which specifically calculates a weighted score based on the feature category label and selects the category with the highest score as the category of the merged patch.

6. The cross-grid patch adaptive merging method for gridded remote sensing interpretation according to claim 1, characterized in that, The preset threshold is dynamically configured based on different land cover types.

7. The cross-grid patch adaptive merging method for gridded remote sensing interpretation according to claim 4, characterized in that, The concave hull algorithm is the Alpha Shape algorithm, which controls the fineness of the contour through the scale parameter α, and the convex hull algorithm is the Convex Hull algorithm.

8. A cross-grid patch adaptive merging system for gridded remote sensing interpretation, characterized in that, include: The first processing module is used to perform preprocessing to expand the interpretation range of the interpretation grid of the remote sensing image; The second processing module is used to obtain multiple patches to be merged based on the interpretation grid after the expansion preprocessing. Each patch includes geometric polygon data and attribute information, and the attribute information includes land cover category and confidence level. The third processing module is used to calculate the spatial intersection area of ​​two spatially adjacent patches, calculate the overlap degree based on the minimum value of the spatial intersection area and the area of ​​the two patches, compare the overlap degree with a preset threshold, and if the overlap degree is greater than or equal to the preset threshold, determine that the two patches should be merged. The fourth processing module is used to construct an undirected graph with patches as nodes and merging relationships as edges based on all the determination results that should be merged. The merging relationship graph is traversed to identify all connected components, and each connected component represents a set of patches to be merged. The fifth processing module is used to merge the geometric vertices of all patches within each connected component into a unified vertex set, and apply a contour reconstruction algorithm to the unified vertex set to generate a merged polygon representing the overall geometric shape. The sixth processing module is used to determine the land cover category of the merged patch based on the attribute information of each patch in the connected component through a voting mechanism, and to take the maximum confidence value of each patch in the connected component as the confidence value of the merged patch. The seventh processing module is used to output the merged patches.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the cross-grid patch adaptive merging method for gridded remote sensing interpretation as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the cross-grid patch adaptive merging method for gridded remote sensing interpretation as described in any one of claims 1 to 7.

Images