Natural resource illegal activity identification method and system based on image recognition

Abstract

This invention relates to the field of natural resource monitoring technology, specifically to a method and system for identifying illegal activities related to natural resources based on image recognition. The method includes the following steps: extracting the reflectance of remote sensing image bands and labeling land surface categories by comparing with preset values; generating a pixel distribution map; identifying connected regions of the same type; extracting boundary pixels and direction sequences to determine the outline of land features; calculating the distance between pixels and boundary lines and matching overlapping segments; obtaining a set of boundary adjacent outlines; and combining center distance sequences and adjacency matching logic to achieve accurate identification of discrete extended outline regions. This invention improves the accuracy of labeling complex land surface categories through multi-band reflectance and index construction; enhances the continuity of land feature outline boundaries by combining connectivity identification and boundary direction sequences; and achieves a fine depiction of resource boundary changes by utilizing boundary adjacency calculation and center distance analysis, effectively identifying the overall spatial correlation of discrete land features.

Description

Technical Field

[0001] This invention relates to the field of natural resource monitoring technology, and in particular to a method and system for identifying illegal activities related to natural resources based on image recognition. Background Technology

[0002] The natural resource monitoring system covers elements such as land, minerals, forests and grasslands, and water areas. Core technologies include remote sensing-based resource extraction, surface change identification, and illegal activity monitoring. This field utilizes remote sensing image interpretation, spatial data overlay, and feature recognition, combined with land classification and administrative boundary data, to systematically analyze the spatial distribution and utilization of resources, supporting administrative supervision of land use, mineral resources, and ecological protection.

[0003] This application extracts the morphological features and spatial relationships of ground features from remote sensing imagery, generates vector contour information of these features based on classification rules, and its core logic lies in spatially overlaying and analyzing the identified vector contours with natural resource management boundary data. By comprehensively utilizing the coverage, adjacency, and distribution relationships between the vector contours and management boundaries, and combining area thresholds and spatial displacement features, it achieves automated identification and precise marking of suspected illegal areas. Through multi-dimensional spatial logic discrimination, the accuracy and continuity of identifying illegal activities related to natural resources are improved.

[0004] Existing natural resource monitoring technologies largely rely on macroscopic image interpretation and static boundary overlay, which has the following drawbacks: First, the identification accuracy is insufficient. Using the overall characteristics of the region as the basis for judgment makes it difficult to distinguish heterogeneous landforms with similar spectral characteristics (such as bare soil and built-up areas), leading to surface classification errors. Second, the boundary topology is lacking. It focuses on judging the coverage relationship while ignoring the subtle adjacency distance and spatial orientation relationship between the outline of landforms and the management boundary, making it difficult to identify the risk of edge encroachment. Third, it lacks spatial correlation analysis capabilities, making it impossible to identify the central evolution trend of discrete and fragmented changes, making it difficult to capture regular illegal expansion behavior and restricting regulatory efficiency. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a method and system for identifying illegal activities related to natural resources based on image recognition. The technical solution is as follows: A method for identifying illegal activities related to natural resources based on image recognition includes the following steps: S1: Obtain the row and column numbers of pixels in the remote sensing image raster data, extract the reflectance values ​​of three bands, form the vegetation index distribution and water body index distribution, compare them with the preset standard values, mark the pixel surface category, and generate a surface type pixel distribution map by recording the corresponding pixel spatial coordinates. S2: Based on the distribution map of land surface type pixels, identify the connectivity of pixels of the same category to form pixel connectivity regions, and arrange the outer pixels of each pixel connectivity region to form a boundary pixel sequence. Extract coordinate and direction changes to generate boundary pixel spatial coordinate sequence and boundary direction record sequence, and combine with numbering to obtain the land feature outline boundary pixel sequence. S3: Connect the vector boundary nodes into a resource boundary line segment sequence, measure the distance between the spatial coordinates of the boundary cells and the resource boundary line segment sequence and record it as the boundary adjacency distance, compare it with the preset adjacency distance standard value to form a set of adjacent cell spatial coordinates, record the direction of change of contour space and distance and identify the corresponding overlapping sections, record the contour number and coordinate range of the ground features, and obtain the set of boundary adjacent contours. S4: Based on the pixel sequence of the feature outline boundary, organize it into a set of spatial coordinates of the feature outline center, arrange them into a center spacing sequence according to spatial distance, identify the changes in adjacent distances and record the feature outline number, call the boundary adjacent outline set to perform number matching and record the matching spatial position range, and obtain the discrete extended outline region.

[0006] As a further aspect of the present invention, the identification method further includes S5, specifically: S5: Form a boundary direction recording sequence according to the boundary cell sequence, record the change of two consecutive directions to form a corner node sequence and mark the spatial grid unit, combine the discrete extended contour region with the boundary adjacent contour set to match the number and mark the spatial grid unit, and obtain the distribution map of suspected illegal natural resource areas.

[0007] As a further aspect of the present invention, the step of obtaining S1 is as follows: S1-1: Obtain the row and column numbers of pixels in the remote sensing image raster data. Collect the red band reflectance, near-infrared band reflectance, and short-wave infrared band reflectance corresponding to each pixel. For the same pixel location, read the near-infrared band reflectance and red band reflectance, perform difference relationship processing, and add the two to form the band sum value. Perform ratio relationship processing to form a vegetation index record. At the same time, read the near-infrared band reflectance and short-wave infrared band reflectance, perform difference relationship processing, and add the two to form the band sum value. Perform ratio relationship processing to form a water body index record. Organize the index records corresponding to each pixel location according to the pixel row and column number order to generate a pixel index distribution matrix. S1-2: Based on the pixel index distribution matrix, read the vegetation index record and water index record of each pixel location row by row, obtain the vegetation index discrimination standard and the water index discrimination standard, compare the vegetation index record and water index record of each pixel location with the vegetation index discrimination standard and the water index discrimination standard respectively, assign the corresponding pixel to the land category code according to the two sets of comparison relationship, and sort them according to the pixel row number and column number order to obtain the land category code array; S1-3: Based on the land surface category coding array, read the row number and column number of each pixel to obtain the row direction resolution, column direction resolution, and image start coordinates in the remote sensing image raster spatial reference parameters. Using the row direction resolution, column direction resolution, and image start coordinates, convert the pixel row number into the vertical spatial coordinates of the pixel and the pixel column number into the horizontal spatial coordinates of the pixel. The vertical spatial coordinates and horizontal spatial coordinates of the pixel together form the pixel spatial coordinates. Associate and organize the pixel spatial coordinates with the corresponding land surface category codes to obtain the land surface type pixel distribution map.

[0008] As a further aspect of the present invention, the step of obtaining S2 is as follows: S2-1: Based on the pixel distribution map of the land surface type, read the pixel row number, pixel column number and land surface category code, detect the corresponding pixel category code of the upper and lower position and the left and right position of each pixel and make a consistency judgment. When the category codes of adjacent pixels are consistent, the corresponding pixels are assigned to the same connected set. At the same time, each connected set is assigned a region number and sorted and recorded according to the pixel row and column order to obtain the pixel connected region number sequence. S2-2: Read the cell row number and cell column number corresponding to each region number one by one according to the cell connected region number sequence, detect the four neighboring positions of each cell, and when there are different region numbers in the adjacent positions, mark the cell as an outer cell, and arrange and record it according to the cell row number order and column number change order. At the same time, read the row number change direction and column number change direction between adjacent outer cells and make a direction determination to obtain the boundary cell direction arrangement sequence. S2-3: Call the boundary cell direction arrangement sequence to read the row number and column number of the outer cells one by one, obtain the row direction resolution, column direction resolution and image starting coordinates in the remote sensing image raster spatial reference parameters, perform spatial position conversion based on the correspondence between cell row number and column number to form cell spatial coordinates, read the row and column change direction between adjacent outer cells and record the direction, and organize the cell spatial coordinates, direction records and area numbers to obtain the ground feature outline boundary cell sequence.

[0009] As a further aspect of the present invention, the step of obtaining S3 is as follows: S3-1: Read the horizontal and vertical coordinates of the vector boundary nodes and connect adjacent nodes in sequence to form a resource boundary line segment sequence. At the same time, read the spatial coordinates of the boundary pixels in the feature outline boundary pixel sequence, measure the spatial distance between the boundary pixel spatial coordinates and the starting and ending coordinates of the resource boundary line segments, record the boundary adjacency distance, obtain the adjacency distance standard threshold and judge the numerical relationship with the boundary adjacency distance, filter the boundary pixel spatial positions that meet the adjacency conditions and organize them in pixel order to generate the adjacent pixel spatial distribution interval. S3-2: Based on the spatial distribution interval of adjacent pixels, read the spatial coordinate order of corresponding boundary pixels one by one, detect the horizontal and vertical spatial change directions of adjacent boundary pixels and record the directions, and at the same time read the arrangement order of adjacent distances of the corresponding boundaries and detect the direction of change of adjacent distances. Make positional correspondence judgment between the spatial change direction and the distance change direction. When the two types of directional relationships are consistent, record the continuous segment of the corresponding boundary pixels, and organize the corresponding feature outline number and the spatial coordinate range of boundary pixels to obtain the outline adjacent segment sequence. S3-3: Call the contour adjacent segment sequence to read the corresponding land feature contour number and boundary cell spatial coordinate range of each continuous segment, merge and organize the continuous segments according to the contour number, and at the same time read and record the number of segments corresponding to each contour number and the spatial arrangement order, associate and identify the contour number and the boundary cell spatial coordinate range, and summarize and organize them according to the contour number order to form a boundary adjacent contour set.

[0010] As a further aspect of the present invention, the step of obtaining S4 is as follows: S4-1: Based on the boundary cell sequence of the ground feature outline, read the spatial coordinates of the boundary cells corresponding to each outline number, and perform centralized position identification on the set of horizontal and vertical coordinates of the boundary cells with the same outline number. Summarize and record the spatial positions of each boundary cell according to the horizontal and vertical positions and establish the spatial coordinates of the outline center. At the same time, organize and label the spatial coordinates of each outline center according to the outline number order and form a coordinate set record to obtain the set of spatial coordinates of the outline center. S4-2: Based on the set of spatial coordinates of the center of the contour, read the spatial coordinates of the center of the adjacent contour one by one, detect the changes in the horizontal distance and the vertical distance between the adjacent center spatial coordinates and record the distance relationship. At the same time, sort the distance relationship according to the arrangement order of the center spatial coordinates and judge the changes in the adjacent distance relationship. When the adjacent distance relationship changes continuously, record the arrangement order of the corresponding contour number and generate a sequence of center spacing change segments. S4-3: Call the sequence of center spacing change segments, read the corresponding contour number arrangement order, and at the same time read the contour number and boundary cell spatial coordinate range in the boundary adjacent contour set. Match the two sets of contour numbers. When the contour number relationship is consistent, record the corresponding boundary cell spatial coordinate range and summarize the spatial position. Record the contour number and corresponding spatial coordinate range in a unified manner to obtain the discrete extended contour region.

[0011] As a further aspect of the present invention, the step of obtaining S5 is as follows: S5-1: Read the spatial coordinates of each boundary cell in the boundary cell sequence of the ground feature outline, and detect the lateral and longitudinal position change directions of adjacent cells according to the boundary cell arrangement order in the boundary cell direction arrangement sequence and record the direction markings. At the same time, continuously record and organize each direction marking according to the cell arrangement order, establish the boundary cell direction change arrangement relationship, and summarize and record according to the cell order to generate a boundary direction record sequence. S5-2: Based on the boundary direction record sequence, read the records of two adjacent directions and detect the direction change relationship. When there is a change in the adjacent direction records, read the spatial coordinates of the corresponding boundary cell and record the corner node identification. At the same time, obtain the spatial grid division parameters and mark the grid unit according to the grid position corresponding to the boundary cell spatial coordinates. Sequentially organize and record the spatial grid unit positions corresponding to each corner node in the corner node identification record to obtain the corner node grid position sequence. S5-3: Call the corner node grid position sequence, read the spatial grid unit position corresponding to each corner node, and at the same time read the contour number and spatial coordinate range in the discrete extended contour area and the contour number information in the boundary adjacent contour set. Match the two sets of contour numbers. When the contour numbers are consistent, record the corresponding spatial grid unit position and perform spatial position sorting and marking. Based on the grid unit position, summarize and record the regional range to obtain the distribution map of suspected illegal natural resource areas.

[0012] As a further aspect of the present invention, the spatial location conversion adopts a coordinate mapping model based on the spatial reference parameters of remote sensing image raster. This model determines the geographic spatial coordinates corresponding to the pixel by combining the row and column numbers of the pixel with the row and column resolution of the image and the starting coordinates of the image and performing a linear transformation.

[0013] A natural resource violation identification system based on image recognition, the system comprising the following modules: Image acquisition module: acquires remote sensing image raster data and spatial location index, extracts multispectral band reflectance records, maps spatial location index to pixel spatial coordinates according to spatial resolution parameters, fuses pixel spatial coordinates and multispectral band reflectance records to construct image pixel information set, and generates remote sensing pixel spectral record set; Surface identification module: Based on the remote sensing pixel spectral record set, extract spectral reflectance features, establish multi-dimensional surface feature index markers, introduce preset environmental feature discrimination criteria to perform category determination, establish pixel surface categories based on multi-dimensional surface feature index markers and associate pixel spatial coordinates, and generate a surface type pixel distribution map; Outline construction module: Extracts spatial coordinates of similar surface pixels from the pixel distribution map of the surface type, identifies connected regions based on topological adjacency and assigns region numbers, extracts the outer location index of the connected regions and arranges them in sequence to construct a boundary pixel sequence, converts it into a boundary pixel spatial coordinate sequence by combining spatial resolution parameters, introduces management area vector nodes to construct a resource boundary line segment sequence, performs spatial position matching between the boundary pixel spatial coordinate sequence and the resource boundary line segment sequence, and generates a feature outline boundary pixel sequence by combining region numbers; Boundary Adjacency Module: Extracts the spatial coordinates of connected region boundary cells based on the feature outline boundary cell sequence, introduces the management area baseline boundary line segment sequence to perform spatial adjacency determination, records the spatial coordinate range of boundary cells that meet the adjacency conditions and associates them with connected region numbers, summarizes the adjacent boundary information and region numbers to construct an outline adjacency annotation set, and generates a boundary adjacency outline set. The illegal activity identification module extracts the spatial coordinate sequence of boundary cells from the set of adjacent boundary contours, extracts the sequence of corner nodes based on the characteristics of cell direction changes, introduces the reference coordinates of the monitoring spatial grid, calibrates the spatial grid unit number corresponding to the corner node, performs spatial matching between the connected region number and the spatial grid unit number, integrates the mapping records of grid unit number and connected region number, and generates a distribution map of suspected illegal areas.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, by extracting the row and column numbers of remote sensing images and multi-band reflectance information, and constructing the distribution of vegetation index and water index, pixel-level surface category marking and spatial coordinate association are achieved, improving the accuracy of surface type identification. Subtle differences between different land features in complex surface environments can be more stably distinguished. By identifying connected regions through the connectivity of pixels of the same category and extracting the outer boundary pixel sequence, while recording boundary direction change information, the expression of land feature outline boundaries is more continuous, and the ability to identify complex edge structures is enhanced. By measuring the distance between boundary pixels and resource boundary line segments and identifying adjacency relationships and overlapping segments, surface changes near resource boundaries can be more finely depicted. By extracting the center coordinates of land feature outlines and analyzing the changes in center spacing, the spatial relationship structure between discretely distributed land features is expressed, and scattered change areas can form an overall extended area identification result. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0017] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0018] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0019] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0020] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0021] Please see Figure 1 This invention provides a technical solution: a method for identifying illegal activities related to natural resources based on image recognition, comprising the following steps: S1: Acquire raster data of remote sensing images for natural resource monitoring, read the row and column indices of each pixel in the image raster structure, and extract the reflectance values ​​of the red band, near-infrared band, and short-wave infrared band. Normalize the reflectance values ​​of the three types of bands at the same pixel location to identify differences and form vegetation index distribution and water body index distribution. Compare the vegetation index distribution and water body index distribution with the preset vegetation index discrimination standard value and the preset water body index discrimination standard value respectively, and mark the pixel surface category. Record the pixel spatial coordinates corresponding to the pixel row and column numbers to generate a surface type pixel distribution map. S2: Based on the pixel distribution map of land surface type, read the row number, column number and corresponding spatial coordinates of each pixel, and at the same time read the land surface category label of the pixel. For pixels of the same land surface category, perform connectivity identification according to the adjacent relationship of row number and column number to form a pixel connectivity region. Read the outer pixels of each pixel connectivity region and arrange them in clockwise order to form a boundary pixel sequence. Read the spatial coordinates of each pixel row number and pixel column number in the boundary pixel sequence and record them to form a boundary pixel spatial coordinate sequence. At the same time, read the changing direction of the row number and the changing direction of the column number of adjacent boundary pixels to form a boundary direction recording sequence. Organize the boundary pixel spatial coordinate sequence and the boundary direction recording sequence and record the connectivity region number to obtain the land feature outline boundary pixel sequence. S3: Obtain the spatial coordinates of the vector boundary nodes of the natural resource management area and connect them in the order of the nodes to form a resource boundary line segment sequence. Read the spatial coordinate sequence of the boundary pixels of the land feature outline based on the pixel sequence of the land feature outline. Measure the spatial distance between the boundary pixel spatial coordinates and the resource boundary line segment sequence and record it as the boundary adjacency distance. Compare the boundary adjacency distance with the preset adjacency distance standard value and filter to form a set of adjacent pixel spatial coordinates. Read the changes in the spatial position of adjacent pixels based on the set of adjacent pixel spatial coordinates and record the direction of the spatial change of the outline. At the same time, read the changes in the boundary adjacency distance according to the order of the boundary pixels and record the direction of the distance change. Match the direction of the spatial change of the outline with the direction of the distance change and identify overlapping sections. Record the land feature outline number and the spatial coordinate range of the boundary pixels corresponding to the overlapping sections to obtain the set of adjacent boundary outlines. S4: Read the spatial coordinate sequence of the boundary pixels of the ground features based on the boundary pixel sequence of the ground features and organize them into a set of spatial coordinates of the center of the ground features. Read the spatial coordinates of the center of the ground features and arrange them in order of spatial distance to form a sequence of center spacing. Continuously identify the relationship between adjacent distance changes in the center spacing sequence and record the corresponding ground feature numbers. At the same time, call the set of adjacent boundary contours to read the ground feature numbers, perform number matching, and record the spatial location range of the matched ground feature contours to obtain the discrete extended contour region. S5: Based on the boundary cell sequence of ground features, read the spatial coordinate sequence of the boundary cells and form a boundary direction record sequence according to the order of the boundary cells. Read two consecutive direction records in the boundary direction record sequence and identify the direction change. At the same time, record the cells connecting the two direction records to form a bend node sequence. Based on the spatial grid cell division, read the spatial coordinates of the cells in the bend node sequence and mark the corresponding spatial grid cells. At the same time, call the discrete extended contour region and the boundary adjacent contour set to read the ground feature contour number for spatial location matching. Mark the spatial grid cell where the matched ground feature contour is located to obtain the distribution map of suspected illegal areas of natural resources.

[0022] Example 1 This embodiment is used to achieve preliminary identification and spatial positioning of land surface categories in remote sensing images. Specifically, the steps for obtaining S1 are as follows: S1-1: Obtain the row and column numbers of pixels in the remote sensing image raster data. Collect the red band reflectance, near-infrared band reflectance, and short-wave infrared band reflectance corresponding to each pixel. For the same pixel location, read the near-infrared band reflectance and red band reflectance, perform difference relationship processing, and add the two to form the band sum value. Perform ratio relationship processing to form a vegetation index record. At the same time, read the near-infrared band reflectance and short-wave infrared band reflectance, perform difference relationship processing, and add the two to form the band sum value. Perform ratio relationship processing to form a water body index record. Organize the index records corresponding to each pixel location according to the pixel row and column number order to generate a pixel index distribution matrix. Obtain the row and column numbers of pixels from the raster data of the remote sensing image. Extract the gray values ​​of specific bands from the multispectral remote sensing image data structure and convert them into reflectance data. Define the reflectance parameters for the red band, near-infrared band, and short-wave infrared band as follows: , and Select a single pixel with row number 100 and column number 200 in the image as an example, and read the red band reflectance of that point. The reflectance in the near-infrared band is 0.12. The reflectance in the shortwave infrared band is 0.58. The value is 0.08, and vegetation index recording is performed for this pixel. The calculation process is achieved by calling... and The numerical values ​​are obtained by performing interpolation. Simultaneously, the summation operation is performed on these two reflectivity values ​​to obtain... The above difference With band sum Perform ratio calculation to obtain Calculation results The vegetation index record value marked for this pixel is used to calculate the water index record for the same pixel location, calling... and The numerical values ​​are obtained by performing interpolation. Simultaneously, the summation operation is performed on these two reflectivity values ​​to obtain... The above difference With band sum Perform ratio calculation to obtain Calculation results The water index record value of the pixel is marked. The reflectance of each pixel in three bands is read in a loop from the first row and first column of the pixel to the last row and last column of the pixel. The above difference and summation ratio calculation logic is executed one by one to obtain the vegetation index and water index value pairs under each row and column index. The index value pairs of all pixels are filled into the preset value array according to the row and column topology relationship to generate the pixel index distribution matrix.

[0023] S1-2: Based on the pixel index distribution matrix, read the vegetation index record and water index record of each pixel location row by row, obtain the vegetation index discrimination standard and the water index discrimination standard, compare the vegetation index record and water index record of each pixel location with the vegetation index discrimination standard and the water index discrimination standard respectively, assign the corresponding pixel to the land category code according to the two sets of comparison relationship, and sort them according to the pixel row number and column number order to obtain the land category code array; Based on the pixel index distribution matrix, vegetation index records and water index records for each pixel location are read row by row. A discrimination standard for vegetation index is pre-defined based on historical experience data and ground feature band response characteristics in natural resource monitoring scenarios. The water quality index is set at 0.3. The value is 0.2. Extract the pixel index record from row 50 and column 80 of the matrix. Read the vegetation index record value at that location as 0.45 and the water index record value as 0.05. Compare the vegetation index record value of 0.45 with the discrimination criteria. Perform numerical comparisons and determine The logic holds true, and the water index record value of 0.05 is compared with the discrimination standard. Perform numerical comparisons and determine The logic holds true. Based on the above comparison results, the land surface category is classified and mapped. When the vegetation index is greater than its standard value and the water index is less than its standard value, the land surface category code "1" is assigned to represent vegetation. If the vegetation index of another pixel is 0.1 and the water index is 0.6, the classification is determined by numerical comparison. and If the index is less than the corresponding standard value, the surface category code "2" is assigned to represent water bodies. If both indices are less than the corresponding standard value, the code "0" is assigned to represent bare land or built-up areas. The entire pixel index distribution matrix is ​​iterated through, and the above logical judgment process is performed on the index values ​​in each row and column cell. The determined value code is filled into the null value array of the corresponding row and column index to obtain the surface category code array.

[0024] S1-3: Based on the land surface category coding array, read the row number and column number of each pixel to obtain the row direction resolution, column direction resolution, and image start coordinates in the remote sensing image raster spatial reference parameters. Using the row direction resolution, column direction resolution, and image start coordinates, convert the pixel row number into the vertical spatial coordinates of the pixel and the pixel column number into the horizontal spatial coordinates of the pixel. The vertical spatial coordinates and horizontal spatial coordinates of the pixel together form the pixel spatial coordinates. Associate and organize the pixel spatial coordinates with the corresponding land surface category codes to obtain the land surface type pixel distribution map. Based on the surface category coding array, the pixel row number and pixel column number are read pixel by pixel. Spatial reference parameters are retrieved from the metadata header file of the remote sensing image to obtain the starting coordinates of the upper left corner of the image and set the starting horizontal coordinate. 450,000 meters, starting ordinate The length is 3,400,000 meters, and the reading resolution is in the line direction. 2 meters, reading column direction resolution It is 2 meters long, with row numbers in the coded array. , column number The pixel is used as a calculation instance to perform vertical spatial coordinate calculation. The conversion logic involves multiplying the row number 500 by the row direction resolution of 2 meters to obtain 1000 meters, and then calculating from the initial ordinate... Subtract the product from the middle, that is, perform the operation. Meters, the vertical spatial coordinate of the pixel is obtained as 3399000 meters, then the horizontal spatial coordinate is executed. The conversion logic involves multiplying column number 300 by the column direction resolution of 2 meters to obtain 600 meters, and then dividing this by the initial horizontal coordinate. Performing addition, that is, performing The horizontal spatial coordinates of the pixel are calculated to be 450600 meters. The calculated horizontal and vertical coordinates are then combined to form a coordinate pair. Retrieve the encoded array The land surface category code corresponding to the location is "1". The coordinate pairs are mapped and associated with the code values ​​one by one. All row and column nodes are traversed and the above process of adding or subtracting the row and column numbers and physical lengths is repeated. All land surface category information with geographic attributes is summarized in grid form to obtain the land surface type pixel distribution map.

[0025] Example 2 Based on the land surface type pixel distribution map obtained in Example 1, this example is used to identify spatially connected regions with the same land surface category and extract their outline boundaries. The specific acquisition steps of S2 are as follows: S2-1: Based on the pixel distribution map of land surface type, read the pixel row number, pixel column number and land surface category code, detect the corresponding pixel category code of the upper and lower position and the left and right position of each pixel and make a consistency judgment. When the category codes of adjacent pixels are consistent, the corresponding pixels are assigned to the same connected set. At the same time, each connected set is assigned a region number and sorted and recorded according to the pixel row and column order to obtain the pixel connected region number sequence. Based on the pixel distribution map of land surface type, the pixel row number, pixel column number, and land surface category code are read, and the coordinates of the currently processed pixel are set as follows. And its surface category code is Using this point as the center, retrieve the adjacent pixels above it. , adjacent pixels below Left adjacent pixels and adjacent pixels on the right The encoded value, if A value of "1" represents vegetation. The codes of the four neighboring locations are extracted for numerical equality determination. If the right-hand pixel code... If both are "1", the logical result is considered true, and... and The row and column indices are combined into the current temporary set. Then, a recursive four-neighbor search is performed on all members of the set, retrieving the pixel with row number 10 and column number 15 in the surface category coding array, whose code is "2" representing water, and detecting the area below it. If the pixel code is also "2", then the two points are associated and detection continues. right side If the code is "0", the logic is considered false, and no inclusion action is performed. Through the above-mentioned equal value comparison action, all spatially adjacent pixel points with equal code values ​​are defined as an independent connected cluster. The first identified vegetation connected cluster is assigned a region number. Assign a region number to a neighboring water body connectivity cluster. The row number and column number of each pixel are matched with the assigned row number and column number. Numbering is used to bind fields, from row number 1 to... Column numbers from 1 to All records are sorted in ascending order to obtain the sequence of cell connected region numbers.

[0026] S2-2: Read the cell row number and cell column number corresponding to each region number one by one according to the cell connected region number sequence, detect the four neighboring positions of each cell, when there are different region numbers in the adjacent positions, mark the cell as an outer cell, and arrange and record them according to the cell row number order and column number change order. At the same time, read the row number change direction and column number change direction between adjacent outer cells and make direction determination to obtain the boundary cell direction arrangement sequence. Based on the pixel connected region numbering sequence, read the corresponding pixel row number and pixel column number for each region number. For the connected region with region number 101, extract all the pixel points it contains. Perform boundary attribute determination on the pixel with row number 100 and column number 200, and retrieve the four neighboring positions of that pixel. , , , The region number in the sequence, if retrieved If the region number is 201 or 0 (representing the background), then it is determined that there is a heterogeneous number at that adjacent location. Mark the cells as outer perimeter cells, collect all cell points within area number 101 that satisfy the discrimination logic, determine the starting cell as the point with the smallest row number and column number in the area, perform chained indexing according to the logic of increasing row number and changing column number within the same row, and read the current outer perimeter cell. With subsequent peripheral pixels The coordinate difference, execute and Subtraction operation, if the difference pair is The direction record value is determined to be "due east". If the difference pair is... Then the direction record value is determined to be "due south". If the difference pair is... The direction record value is determined to be "due west". If the difference pair is... The direction record value is determined to be "due north". The above step-by-step direction extraction is performed on the edge of the discrete patch of illegally occupied farmland. The spatial order of each peripheral cell is mapped and connected with the corresponding direction description field to obtain the direction arrangement sequence of the boundary cells.

[0027] S2-3: Call the boundary cell orientation sequence to read the row number and column number of the outer cells one by one, obtain the row direction resolution, column direction resolution and image start coordinates in the remote sensing image raster spatial reference parameters, perform spatial position conversion based on the correspondence between cell row number and column number to form cell spatial coordinates, read the row and column change direction between adjacent outer cells and record the direction, and organize the cell spatial coordinates, direction records and area numbers to obtain the feature outline boundary cell sequence; By calling the boundary cell orientation sequence to read the row and column numbers of the outer cells one by one, the row orientation resolution is extracted from the multispectral raster metadata. Meter and column direction resolution Meters, obtain the reference x-coordinate of the upper left corner of the image. and reference ordinate For the row number in the outer pixel sequence , column number is Spatial coordinate calculation is performed on the points. Spatial location conversion adopts a coordinate mapping model based on the spatial reference parameters of remote sensing image raster. This model determines the geospatial coordinates of the pixel by combining the row and column numbers of the pixel with the row and column resolution of the image and the starting coordinates of the image through linear transformation. Multiplying the row number 150 with the row direction resolution 2.0 yields 300.0. Subtracting 300.0 from the reference ordinate 3000000.0 gives... Multiply column number 250 by column direction resolution 2.0 to get 500.0, then add it to the baseline x-coordinate 500000.0 to get... , obtain pixel space coordinate pairs Simultaneously, it retrieves the direction determination result corresponding to that point in the direction arrangement sequence, such as the "due east" direction record, and outputs the calculated spatial coordinate pair, the corresponding direction string, and the region number to which the point belongs. Triple encapsulation is performed, and the coordinate transformation and information merging operations are performed on all peripheral pixels in the order of the boundary index generated by clockwise tracing. The edge information of each independent connected region is structured and summarized to obtain the boundary pixel sequence of the ground features.

[0028] Example 3 Based on the feature outline boundary pixel sequence obtained in Example 2, this example is used to identify feature outlines adjacent to the natural resource management boundary. Specifically, the acquisition step S3 is as follows: S3-1: Read the horizontal and vertical coordinates of the vector boundary nodes and connect adjacent nodes in sequence to form a resource boundary line segment sequence. At the same time, read the spatial coordinates of the boundary pixels in the feature outline boundary pixel sequence, measure the spatial distance between the boundary pixel spatial coordinates and the starting and ending coordinates of the resource boundary line segments, record the boundary adjacency distance, obtain the adjacency distance standard threshold and judge the numerical relationship with the boundary adjacency distance, filter the spatial positions of boundary pixels that meet the adjacency conditions and organize them in pixel order to generate the spatial distribution range of adjacent pixels. Read the horizontal and vertical coordinates of the vector boundary nodes and connect adjacent nodes sequentially to form a resource boundary line segment sequence. Extract the spatial coordinate pair of the first outermost pixel from the feature outline boundary pixel sequence. for Simultaneously retrieve the starting coordinates of a certain line segment in the resource boundary line segment sequence. for coordinates of the endpoint for Perform boundary adjacency distance The measurement action calls the pixel coordinates and the coordinates of the two endpoints of the line segment to perform vertical projection distance calculation. Calculate the square of the line segment length ; Calculate the projection ratio of the vector formed by the pixel point and the starting point onto the line segment vector. ; determination Value at Within the interval, select the coordinates of the projection point. Performing Euclidean distance calculation yields rice; Set the standard threshold for adjacency distance The value is 10.0 meters. This value is set with reference to five times the width of the remote sensing image pixel resolution of 2.0 meters. A value magnitude comparison is then performed to determine... If the logic is true, then the pixel space coordinates will be... It is determined that the adjacency condition is met; If the distance calculated by another pixel is 15.0 meters, then... If the result is false, an exclusion action will be performed.

[0029] Iterate through all coordinate points in the boundary cell sequence, extract and arrange the coordinates of all true cells according to their index positions in the original sequence, and generate the spatial distribution range of adjacent cells.

[0030] S3-2: Based on the spatial distribution interval of adjacent pixels, read the spatial coordinate order of corresponding boundary pixels one by one, detect the horizontal and vertical spatial change directions of adjacent boundary pixels and record the directions, and at the same time read the arrangement order of adjacent distances of the corresponding boundaries and detect the direction of change of adjacent distances. Make positional correspondence judgment between the spatial change direction and the distance change direction. When the two types of directional relationships are consistent, record the continuous segment of the corresponding boundary pixels, and organize the corresponding feature outline number and the spatial coordinate range of boundary pixels to obtain the outline adjacent segment sequence. Based on the spatial distribution interval of adjacent pixels, the spatial coordinates of corresponding boundary pixels are read sequentially, and the coordinates of two consecutive pixels within the interval are retrieved. and : Perform coordinate subtraction to obtain the horizontal change. ; Obtaining longitudinal change ; The horizontal change direction is determined to be "flat", and the vertical change direction is determined to be "positive increase". (The data is then read.) Corresponding boundary adjacency distance Rice and corresponding distance Meters, perform distance subtraction operation If the direction of distance change is determined to be "flat", then a consistency judgment action for two types of directions is performed. When the spatial coordinates move in the positive direction and the adjacent distance remains unchanged, it is determined that the boundary segment and the resource boundary segment are in a parallel adjacent state. The set of continuous cells for which this logic holds is recorded, and the corresponding land feature outline number is extracted as 101. The coordinates of its starting cell are then locked. and the coordinates of the terminating cell As the boundary cell spatial coordinate range, if the direction of distance change is subsequently detected to change from "flat" to "increasing", it is determined that the directional relationship is inconsistent. The current segment is then truncated and a new sequence search is started. All cell segments that meet the consistency judgment and their attributes are recorded as entries to obtain the contour adjacent segment sequence.

[0031] S3-3: Call the contour adjacent segment sequence to read the corresponding land feature contour number and boundary cell spatial coordinate range of each continuous segment, merge and organize the continuous segments according to the contour number, and at the same time read and record the number of segments corresponding to each contour number and the spatial arrangement order, associate and identify the contour number and the boundary cell spatial coordinate range, and summarize and organize them according to the contour number order to form a boundary adjacent contour set. The contour adjacent segment sequence is called to read the corresponding feature contour number and boundary cell spatial coordinate range of each continuous segment. All segment records belonging to number 101 in the sequence are retrieved, and the coordinate range of the first segment is extracted. Extract the coordinate range of the second segment. Perform a merging and sorting operation for segments with the same number. Count the total number of segments corresponding to number 101 as 2. Record their spatial arrangement order as first north-south and then east-west. Perform field association mapping between number 101 and these two sets of spatial coordinate ranges. Then retrieve the segment record for number 102. If this number corresponds to only 1 segment, record the segment count as 1 and extract its corresponding coordinate range. According to the outline numbering from 101 to The numerical ascending order rule is used to structurally merge the number of segments, spatial location orientation, and specific physical coordinate boundary values ​​under each number name to generate a set of boundary adjacency contours.

[0032] Example 4 Based on the boundary adjacency contour set obtained in Example 3, this example is used to identify ground feature contour groups exhibiting a discrete expansion trend. Specifically, the steps for obtaining S4 are as follows: S4-1: Based on the boundary cell sequence of the ground feature outline, read the spatial coordinates of the boundary cells corresponding to each outline number, and perform centralized position identification on the set of horizontal and vertical coordinates of the boundary cells with the same outline number. Summarize and record the spatial positions of each boundary cell according to the horizontal and vertical positions and establish the spatial coordinates of the outline center. At the same time, organize and label the spatial coordinates of each outline center according to the outline number order and form a coordinate set record to obtain the set of spatial coordinates of the outline center. Based on the boundary cell sequence of the ground feature outlines, the spatial coordinates of the boundary cells corresponding to each outline number are read. For the ground feature outline numbered 101, the set of lateral coordinates of all its boundary cells is extracted. With the set of vertical coordinates If the contour contains four boundary points with coordinates as follows: , , , Perform the arithmetic mean calculation of the coordinates in each dimension; Call the horizontal coordinate to execute rice; Call the vertical coordinate to execute rice; Pair the obtained mean Establish the center spatial coordinates of this contour Repeat the above centralized location identification action to process the outline of feature number 102. If the calculated mean value of its boundary coordinates is... Then establish the central coordinates The identified center points of each contour are arranged in the order of their numerical values, such as 101, 102, 103, etc. Each center coordinate is mapped one-to-one with its corresponding feature contour number. All mapped coordinate entries are stored in a specified field of a structured database to obtain the set of spatial coordinates of the contour centers.

[0033] S4-2: Based on the set of spatial coordinates of the center of the contour, read the spatial coordinates of the center of the adjacent contour one by one, detect the changes in the horizontal and vertical distances between the adjacent center spatial coordinates and record the distance relationship. At the same time, sort the distance relationship according to the arrangement order of the center spatial coordinates and judge the changes in the adjacent distance relationship. When the adjacent distance relationship changes continuously, record the arrangement order of the corresponding contour number and generate a sequence of center spacing change segments. Based on the spatial coordinate set of the contour center, the spatial coordinates of adjacent contour centers are read sequentially to extract the center coordinates of number 101. Center coordinates of number 102 ; Perform lateral distance difference calculation rice; Perform vertical distance difference calculation rice; Record the straight-line distance between the adjacent centers. rice; Continue reading the straight-line distance between the centers of numbers 102 and 103. The system calculates various distance values ​​and forms a numerical sequence according to the index of the center point. It then performs a judgment action on changes in adjacent distance relationships and sets a threshold for continuous change judgment deviation. The value is 2.0 meters, which is set with reference to one time the error of the sensor's horizontal positioning accuracy. implement rice; determination If the logical result is true, it is marked as a smooth and continuous change in distance relationship. If the next set of distance differences is also less than 2.0 meters, then the contours within this segment are determined to have a spatially distributed regular relationship, and the contour number sequence that satisfies this continuous logic is recorded as follows: When a spacing difference greater than 2.0 meters is detected, it is determined to be a discrete mutation and the current segment recording is terminated. All number combinations that conform to the stationary and continuous law are summarized to obtain the sequence of segments with changing center spacing.

[0034] S4-3: Call the sequence of center spacing change segments, read the corresponding contour number arrangement order, and at the same time read the contour number and boundary cell spatial coordinate range in the boundary adjacent contour set. Match the two sets of contour numbers. When the contour number relationship is consistent, record the corresponding boundary cell spatial coordinate range and summarize the spatial position. Record the contour number and corresponding spatial coordinate range in a unified manner to obtain the discrete extended contour region. The sequence of center spacing variation segments is called to read the corresponding contour number arrangement order, and the contour numbers contained in the sequence are extracted. Simultaneously, the stored contour number indices in the boundary adjacent contour set are retrieved, and an equivalence matching judgment is performed between the two sets of numbers. For number 101, it is found that it is in the center spacing variation sequence and also exists in the boundary adjacent contour set. It is determined that this contour has both boundary adjacent features and spatial distribution pattern features, and the spatial coordinate range of the boundary cell corresponding to this number is extracted. Continue performing the same matching search on number 102. If number 102 also satisfies the dual set existence judgment, extract its coordinate range and perform a union and superposition operation with the coordinate data of 101. Perform physical spatial range aggregation processing on all successfully matched numbers. Describe the coordinates of these interrelated land feature patches that are close to the management boundary in a closed manner according to the minimum bounding rectangle or polygon boundary. Assign a unified discrete extended identification code to the closed spatial region. Associate and store the aggregated latitude and longitude or projected coordinate point set with the identification code to obtain the discrete extended contour region.

[0035] Example 5 Based on the discrete extended contour region obtained in Example 4, this example is used to identify areas suspected of illegal occupation of natural resources. The acquisition step S5 is as follows: S5-1: Read the spatial coordinates of each boundary cell in the boundary cell sequence of the ground feature outline, and detect the lateral and longitudinal position change directions of adjacent cells according to the boundary cell arrangement order in the boundary cell direction arrangement sequence and record the direction markings. At the same time, continuously record and organize each direction marking according to the cell arrangement order, establish the boundary cell direction change arrangement relationship, summarize and record according to the cell order, and generate the boundary direction record sequence. Read the spatial coordinates of each boundary cell in the boundary cell sequence of the ground feature outline, and extract the clockwise arranged cell coordinate sequence for the outline boundary numbered 101. Select index number as cell coordinates With index number adjacent cell coordinates : Perform lateral position change calculation rice; Calculation of vertical position change rice; Set direction judgment logic, if and Then assign the direction identifier "E" to represent moving eastward. and The direction identifier "N" is assigned to represent moving north. For this instance, the direction identifier is determined to be "E", and the index number is read again. The pixel coordinates are The subtraction operation yields a horizontal change of 2.0 and a vertical change of 0. The direction identifier is then determined to be "E". This series of direction identifier codes generated based on the pixel movement sequence is linearly arranged according to the time step to establish the correspondence between each point and its movement vector direction. If the pixel sequence contains 200 coordinate points, a record chain containing 199 direction identifiers is generated. All direction identifier chains corresponding to the contours of the ground features are archived according to the contour number to generate a boundary direction record sequence.

[0036] S5-2: Based on the boundary direction record sequence, read the records of two adjacent directions and detect the direction change relationship. When there is a change in the records of adjacent directions, read the spatial coordinates of the corresponding boundary cell and record the corner node identification. At the same time, obtain the spatial grid division parameters and mark the grid unit according to the grid position corresponding to the spatial coordinates of the boundary cell. Sequentially organize and record the spatial grid unit positions corresponding to each corner node in the corner node identification record to obtain the corner node grid position sequence. Based on the boundary direction record sequence, read the records of two adjacent directions and detect the direction change relationship, extract the contour numbered 101 in the index. Location direction record With index Location direction record The direction is compared and determined. Since "E" and "N" are not equal, the direction record has changed. The spatial coordinates of the connection point cell that caused the change are read. Mark it as a corner node and assign it a node number. Retrieve spatial mesh generation parameters and set mesh cell size. The value is 10.0 meters, set with reference to the minimum typical feature size of illegal buildings within the monitoring area, to obtain the coordinates of the starting reference point for the image. Execute the grid row index where the corner node is located. Operations: calculate ; Execute grid column index Operations, calculations ; Map this node to coordinates Within the grid cell, the above row and column offsets are divided by the grid step size for all direction switching points on the contour boundary. The grid coordinates of all corner points belonging to the same contour number are recorded in order of their position on the boundary to obtain the grid position sequence of the corner nodes.

[0037] S5-3: Call the corner node grid position sequence, read the spatial grid cell position corresponding to each corner node, and at the same time read the contour number and spatial coordinate range in the discrete extended contour area and the contour number information in the boundary adjacent contour set. Match the two sets of contour numbers. When the contour numbers are consistent, record the corresponding spatial grid cell position and perform spatial position sorting and marking. Based on the grid cell position, summarize and record the area range to obtain the distribution map of suspected illegal natural resource areas. The corner node mesh position sequence is called to read the spatial mesh cell position of each corner node, and the contour number is read from the discrete extended contour region. The records and their covered spatial coordinate ranges are recorded. Simultaneously, a list of contour numbers adjacent to the natural resource management boundary is retrieved from the boundary adjacency contour set. Number 101 within the discrete extended region is extracted and matched with numbers in the boundary adjacency set. If number 101 exists in both sets, it is identified as a key monitoring target. All corner node grid positions corresponding to this number are then extracted. , , And combined with the spatial coordinate range under the outline number name The set of all grid cells affected by the violation is defined. Grid cells that meet the dual numbering matching condition are marked with high brightness. Grid patches that are within the same administrative jurisdiction and exhibit discrete expansion characteristics are merged into polygons. The vector boundary nodes and grid density characteristics of the merged grids are recorded. These marked grid cells are stored in multiple nested layers according to the hierarchical structure of the geographic information system to generate a distribution map of suspected illegal natural resource areas.

[0038] Example 6 Please see Figure 2 This embodiment provides a natural resource violation identification system based on image recognition. This system is used in the natural resource violation identification method based on image recognition described in the above embodiment, and includes the following modules: Image acquisition module: acquires remote sensing image raster data and spatial location index, extracts multispectral band reflectance records, maps spatial location index to pixel spatial coordinates according to spatial resolution parameters, fuses pixel spatial coordinates and multispectral band reflectance records to construct image pixel information set, and generates remote sensing pixel spectral record set; Surface identification module: Extract spectral reflectance features based on remote sensing pixel spectral record set, establish multidimensional surface feature index markers, introduce preset environmental feature discrimination criteria to perform category determination, establish pixel surface category based on multidimensional surface feature index markers and associate pixel spatial coordinates, and generate a surface type pixel distribution map; Outline construction module: Extracts spatial coordinates of similar surface pixels from the surface type pixel distribution map, identifies connected regions based on topological adjacency and assigns region numbers, extracts the outer location index of connected regions and arranges them in sequence to construct a boundary pixel sequence, converts it into a boundary pixel spatial coordinate sequence by combining spatial resolution parameters, introduces management area vector nodes to construct a resource boundary line segment sequence, performs spatial position matching between the boundary pixel spatial coordinate sequence and the resource boundary line segment sequence, and generates a feature outline boundary pixel sequence by combining region numbers; Boundary Adjacency Module: Extracts the spatial coordinates of boundary cells of connected regions based on the boundary cell sequence of ground feature outlines, introduces the baseline boundary line segment sequence of the management area to perform spatial adjacency determination, records the spatial coordinate range of boundary cells that meet the adjacency conditions and associates them with the connected region number, summarizes the adjacent boundary information and region number to construct an outline adjacency annotation set, and generates a boundary adjacency outline set. The illegal activity identification module extracts the spatial coordinate sequence of boundary cells from the set of adjacent boundary contours, extracts the sequence of corner nodes based on the characteristics of cell direction changes, introduces the reference coordinates of the monitoring spatial grid, calibrates the spatial grid cell number corresponding to the corner node, performs spatial matching between the connected region number and the spatial grid cell number, integrates the mapping records of grid cell number and connected region number, and generates a distribution map of suspected illegal areas.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying illegal activities related to natural resources based on image recognition, characterized in that, Includes the following steps: S1: Obtain the row and column numbers of pixels in the remote sensing image raster data, extract the reflectance values ​​of three bands, form the vegetation index distribution and water body index distribution, compare them with the preset standard values, mark the pixel surface category, and generate a surface type pixel distribution map by recording the corresponding pixel spatial coordinates. S2: Based on the distribution map of land surface type pixels, identify the connectivity of pixels of the same category to form pixel connectivity regions, and arrange the outer pixels of each pixel connectivity region to form a boundary pixel sequence. Extract coordinate and direction changes to generate boundary pixel spatial coordinate sequence and boundary direction record sequence, and combine with numbering to obtain the land feature outline boundary pixel sequence. S3: Connect the vector boundary nodes into a resource boundary line segment sequence, measure the distance between the spatial coordinates of the boundary cells and the resource boundary line segment sequence and record it as the boundary adjacency distance, compare it with the preset adjacency distance standard value to form a set of adjacent cell spatial coordinates, record the direction of change of contour space and distance and identify the corresponding overlapping sections, record the contour number and coordinate range of the ground features, and obtain the set of boundary adjacent contours. S4: Based on the pixel sequence of the feature outline boundary, organize it into a set of spatial coordinates of the feature outline center, arrange them into a center spacing sequence according to spatial distance, identify the changes in adjacent distances and record the feature outline number, call the boundary adjacent outline set to perform number matching and record the matching spatial position range, and obtain the discrete extended outline region.

2. The method for identifying illegal activities related to natural resources based on image recognition according to claim 1, characterized in that: The identification method further includes S5, specifically: S5: Form a boundary direction recording sequence according to the boundary cell sequence, record the change of two consecutive directions to form a corner node sequence and mark the spatial grid unit, combine the discrete extended contour region with the boundary adjacent contour set to match the number and mark the spatial grid unit, and obtain the distribution map of suspected illegal natural resource areas.

3. The method for identifying illegal activities related to natural resources based on image recognition according to claim 1, characterized in that, The steps for obtaining S1 are as follows: S1-1: Obtain the row and column numbers of pixels in the remote sensing image raster data. Collect the red band reflectance, near-infrared band reflectance, and short-wave infrared band reflectance corresponding to each pixel. For the same pixel location, read the near-infrared band reflectance and red band reflectance, perform difference relationship processing, and add the two to form the band sum value. Perform ratio relationship processing to form a vegetation index record. At the same time, read the near-infrared band reflectance and short-wave infrared band reflectance, perform difference relationship processing, and add the two to form the band sum value. Perform ratio relationship processing to form a water body index record. Organize the index records corresponding to each pixel location according to the pixel row and column number order to generate a pixel index distribution matrix. S1-2: Based on the pixel index distribution matrix, read the vegetation index record and water index record of each pixel location row by row, obtain the vegetation index discrimination standard and the water index discrimination standard, compare the vegetation index record and water index record of each pixel location with the vegetation index discrimination standard and the water index discrimination standard respectively, assign the corresponding pixel to the land category code according to the two sets of comparison relationship, and sort them according to the pixel row number and column number order to obtain the land category code array; S1-3: Based on the land surface category coding array, read the row number and column number of each pixel to obtain the row direction resolution, column direction resolution, and image start coordinates in the remote sensing image raster spatial reference parameters. Using the row direction resolution, column direction resolution, and image start coordinates, convert the pixel row number into the vertical spatial coordinates of the pixel and the pixel column number into the horizontal spatial coordinates of the pixel. The vertical spatial coordinates and horizontal spatial coordinates of the pixel together form the pixel spatial coordinates. Associate and organize the pixel spatial coordinates with the corresponding land surface category codes to obtain the land surface type pixel distribution map.

4. The method for identifying illegal activities related to natural resources based on image recognition according to claim 1, characterized in that, The steps for obtaining S2 are as follows: S2-1: Based on the pixel distribution map of the land surface type, read the pixel row number, pixel column number and land surface category code, detect the corresponding pixel category code of the upper and lower position and the left and right position of each pixel and make a consistency judgment. When the category codes of adjacent pixels are consistent, the corresponding pixels are assigned to the same connected set. At the same time, each connected set is assigned a region number and sorted and recorded according to the pixel row and column order to obtain the pixel connected region number sequence. S2-2: Read the cell row number and cell column number corresponding to each region number one by one according to the cell connected region number sequence, detect the four neighboring positions of each cell, and when there are different region numbers in the adjacent positions, mark the cell as an outer cell, and arrange and record it according to the cell row number order and column number change order. At the same time, read the row number change direction and column number change direction between adjacent outer cells and make a direction determination to obtain the boundary cell direction arrangement sequence. S2-3: Call the boundary cell direction arrangement sequence to read the row number and column number of the outer cells one by one, obtain the row direction resolution, column direction resolution and image starting coordinates in the remote sensing image raster spatial reference parameters, perform spatial position conversion based on the correspondence between cell row number and column number to form cell spatial coordinates, read the row and column change direction between adjacent outer cells and record the direction, and organize the cell spatial coordinates, direction records and area numbers to obtain the ground feature outline boundary cell sequence.

5. The method for identifying illegal activities related to natural resources based on image recognition according to claim 1, characterized in that, The steps for obtaining S3 are as follows: S3-1: Read the horizontal and vertical coordinates of the vector boundary nodes and connect adjacent nodes in sequence to form a resource boundary line segment sequence. At the same time, read the spatial coordinates of the boundary pixels in the feature outline boundary pixel sequence, measure the spatial distance between the boundary pixel spatial coordinates and the starting and ending coordinates of the resource boundary line segments, record the boundary adjacency distance, obtain the adjacency distance standard threshold and judge the numerical relationship with the boundary adjacency distance, filter the boundary pixel spatial positions that meet the adjacency conditions and organize them in pixel order to generate the adjacent pixel spatial distribution interval. S3-2: Based on the spatial distribution interval of adjacent pixels, read the spatial coordinate order of corresponding boundary pixels one by one, detect the horizontal and vertical spatial change directions of adjacent boundary pixels and record the directions, and at the same time read the arrangement order of adjacent distances of the corresponding boundaries and detect the direction of change of adjacent distances. Make positional correspondence judgment between the spatial change direction and the distance change direction. When the two types of directional relationships are consistent, record the continuous segment of the corresponding boundary pixels, and organize the corresponding feature outline number and the spatial coordinate range of boundary pixels to obtain the outline adjacent segment sequence. S3-3: Call the contour adjacent segment sequence to read the corresponding land feature contour number and boundary cell spatial coordinate range of each continuous segment, merge and organize the continuous segments according to the contour number, and at the same time read and record the number of segments corresponding to each contour number and the spatial arrangement order, associate and identify the contour number and the boundary cell spatial coordinate range, and summarize and organize them according to the contour number order to form a boundary adjacent contour set.

6. The method for identifying illegal activities related to natural resources based on image recognition according to claim 1, characterized in that, The steps for obtaining S4 are as follows: S4-1: Based on the boundary cell sequence of the ground feature outline, read the spatial coordinates of the boundary cells corresponding to each outline number, and perform centralized position identification on the set of horizontal and vertical coordinates of the boundary cells with the same outline number. Summarize and record the spatial positions of each boundary cell according to the horizontal and vertical positions and establish the spatial coordinates of the outline center. At the same time, organize and label the spatial coordinates of each outline center according to the outline number order and form a coordinate set record to obtain the set of spatial coordinates of the outline center. S4-2: Based on the set of spatial coordinates of the center of the contour, read the spatial coordinates of the center of the adjacent contour one by one, detect the changes in the horizontal distance and the vertical distance between the adjacent center spatial coordinates and record the distance relationship. At the same time, sort the distance relationship according to the arrangement order of the center spatial coordinates and judge the changes in the adjacent distance relationship. When the adjacent distance relationship changes continuously, record the arrangement order of the corresponding contour number and generate a sequence of center spacing change segments. S4-3: Call the sequence of center spacing change segments, read the corresponding contour number arrangement order, and at the same time read the contour number and boundary cell spatial coordinate range in the boundary adjacent contour set. Match the two sets of contour numbers. When the contour number relationship is consistent, record the corresponding boundary cell spatial coordinate range and summarize the spatial position. Record the contour number and corresponding spatial coordinate range in a unified manner to obtain the discrete extended contour region.

7. The method for identifying illegal activities related to natural resources based on image recognition according to claim 2, characterized in that, The steps for obtaining S5 are as follows: S5-1: Read the spatial coordinates of each boundary cell in the boundary cell sequence of the ground feature outline, and detect the lateral and longitudinal position change directions of adjacent cells according to the boundary cell arrangement order in the boundary cell direction arrangement sequence and record the direction markings. At the same time, continuously record and organize each direction marking according to the cell arrangement order, establish the boundary cell direction change arrangement relationship, and summarize and record according to the cell order to generate a boundary direction record sequence. S5-2: Based on the boundary direction record sequence, read the records of two adjacent directions and detect the direction change relationship. When there is a change in the adjacent direction records, read the spatial coordinates of the corresponding boundary cell and record the corner node identification. At the same time, obtain the spatial grid division parameters and mark the grid unit according to the grid position corresponding to the boundary cell spatial coordinates. Sequentially organize and record the spatial grid unit positions corresponding to each corner node in the corner node identification record to obtain the corner node grid position sequence. S5-3: Call the corner node grid position sequence, read the spatial grid unit position corresponding to each corner node, and at the same time read the contour number and spatial coordinate range in the discrete extended contour area and the contour number information in the boundary adjacent contour set. Match the two sets of contour numbers. When the contour numbers are consistent, record the corresponding spatial grid unit position and perform spatial position sorting and marking. Based on the grid unit position, summarize and record the regional range to obtain the distribution map of suspected illegal natural resource areas.

8. The method for identifying illegal activities related to natural resources based on image recognition according to claim 4, characterized in that, The spatial location conversion adopts a coordinate mapping model based on the spatial reference parameters of remote sensing image raster. This model determines the geospatial coordinates corresponding to the pixel by combining the row and column numbers of the pixel with the row and column resolution of the image and the starting coordinates of the image and performing a linear transformation.

9. A natural resource violation identification system based on image recognition, characterized in that, The system is used in the natural resource violation identification method based on image recognition as described in any one of claims 1-8, and the system includes the following modules: Image acquisition module: acquires remote sensing image raster data and spatial location index, extracts multispectral band reflectance records, maps spatial location index to pixel spatial coordinates according to spatial resolution parameters, fuses pixel spatial coordinates and multispectral band reflectance records to construct image pixel information set, and generates remote sensing pixel spectral record set; Surface identification module: Based on the remote sensing pixel spectral record set, extract spectral reflectance features, establish multi-dimensional surface feature index markers, introduce preset environmental feature discrimination criteria to perform category determination, establish pixel surface categories based on multi-dimensional surface feature index markers and associate pixel spatial coordinates, and generate a surface type pixel distribution map; Outline construction module: Extracts spatial coordinates of similar surface pixels from the pixel distribution map of the surface type, identifies connected regions based on topological adjacency and assigns region numbers, extracts the outer location index of the connected regions and arranges them in sequence to construct a boundary pixel sequence, converts it into a boundary pixel spatial coordinate sequence by combining spatial resolution parameters, introduces management area vector nodes to construct a resource boundary line segment sequence, performs spatial position matching between the boundary pixel spatial coordinate sequence and the resource boundary line segment sequence, and generates a feature outline boundary pixel sequence by combining region numbers; Boundary Adjacency Module: Extracts the spatial coordinates of connected region boundary cells based on the feature outline boundary cell sequence, introduces the management area baseline boundary line segment sequence to perform spatial adjacency determination, records the spatial coordinate range of boundary cells that meet the adjacency conditions and associates them with connected region numbers, summarizes the adjacent boundary information and region numbers to construct an outline adjacency annotation set, and generates a boundary adjacency outline set. The illegal activity identification module extracts the spatial coordinate sequence of boundary cells from the set of adjacent boundary contours, extracts the sequence of corner nodes based on the characteristics of cell direction changes, introduces the reference coordinates of the monitoring spatial grid, calibrates the spatial grid unit number corresponding to the corner node, performs spatial matching between the connected region number and the spatial grid unit number, integrates the mapping records of grid unit number and connected region number, and generates a distribution map of suspected illegal areas.

Images