A method, system, and medium for supplemental cropland restoration assessment

Abstract

The application belongs to the field of cultivated land detection and evaluation, and particularly relates to a cultivated land recovery evaluation method and system and a medium. The advantages of flexibility and short collection period of the unmanned aerial vehicle are utilized to realize accurate mapping of geographic coordinates of multi-source data in cooperation. Specifically, SIFT feature matching algorithm is performed on the unmanned aerial vehicle image and high-resolution remote sensing image, pixel coordinate geographic conversion of the unmanned aerial vehicle image is solved, the cultivated land patch is mapped onto the unmanned aerial vehicle image, and the accuracy of cultivated land range verification is ensured. Meanwhile, the YOLO algorithm is used to perform instance segmentation on the cultivated land patch range mapped by the unmanned aerial vehicle image, and the cultivated land and non-cultivated land in the geographic patch are accurately distinguished, thereby providing solid data support for quantitative evaluation of the cultivated land area. The problems of time lag, insufficient precision and high cost existing in long-term supervision of the cultivated land are effectively solved, and reliable technical support is provided for intelligent monitoring of the cultivated land.

Description

Technical Field

[0001] The present application relates to the technical field of supplementary cultivated land detection and assessment, and in particular to a supplementary cultivated land restoration assessment method, system and medium. Background Art

[0002] Traditional methods for monitoring supplementary cultivated land rely primarily on manual inspections and spot checks, which suffer from long processing times, low efficiency, and susceptibility to human interference, making them inadequate for modern monitoring. In recent years, remote sensing image interpretation has been widely used for land classification and identification, but its practical application still faces numerous limitations. For one thing, commercial remote sensing imagery is expensive to acquire, and due to its long acquisition cycles, the update frequency cannot meet the needs of real-time monitoring. Furthermore, remote sensing imagery is susceptible to factors such as weather, season, and lighting. Furthermore, remote sensing image interpretation typically involves processing large-scale data, resulting in high computational effort and lengthy interpretation times, making it difficult to meet the demands of rapid monitoring. These factors can lead to reduced image stability and consistency, thus compromising the accuracy of cultivated land segmentation.

[0003] In contrast, drones, with their low cost, high frequency, and high resolution, offer greater flexibility in farmland monitoring and are suitable for dynamic monitoring. However, current drone inspections of supplementary farmland rely solely on video streams and manual visual analysis, lacking the full integration of intelligent algorithms for automated identification and analysis. This makes it difficult to effectively detect fraudulent supplementary farmland inspections. Therefore, integrating deep learning and computer vision to transform drone inspections from passive monitoring to intelligent analysis is key to improving regulatory efficiency. Summary of the Invention

[0004] The embodiments of the present application provide a method, system and medium for evaluating the restoration of supplementary cultivated land. Using drones as a carrier and combining feature matching and segmentation algorithms, it can achieve accurate identification and dynamic monitoring of supplementary cultivated land, improve the timeliness and flexibility of monitoring, and solve the problems of monitoring lag and low efficiency in the current supervision of supplementary cultivated land.

[0005] To this end, according to one aspect of the present application, a method for evaluating supplementary cultivated land restoration is provided, comprising the following steps:

[0006] Obtain vector boundary layers and high-resolution remote sensing images of supplementary cultivated land;

[0007] Deploy the drone and navigate to the geometric center coordinates of each supplementary cultivated land patch to collect drone images of the supplementary cultivated land area;

[0008] Use the SIFT feature matching algorithm to extract feature points from high-resolution remote sensing images and UAV images, and calculate and evaluate the homography matrix;

[0009] Use homography matrix to perform perspective transformation to map the supplementary cultivated land patches of high-resolution remote sensing imagery into UAV imagery;

[0010] The cultivated land and non-cultivated land areas within the supplementary cultivated land map are segmented based on the YOLO algorithm, and their geographical areas and proportions are calculated respectively through the homography matrix.

[0011] Optionally, obtaining a vector boundary layer and a high-resolution remote sensing image of supplementary cultivated land specifically includes the following steps:

[0012] Obtain the vector boundary layer of the supplementary cultivated land from the cultivated land business system, and obtain a high-resolution remote sensing image with a resolution better than 1 meter. The size of the high-resolution remote sensing image should cover the entire vector boundary layer of the supplementary cultivated land. If the coordinates of the vector boundary layer of the supplementary cultivated land are inconsistent with those of the high-resolution remote sensing image, perform coordinate conversion on the vector boundary layer to make it consistent with the coordinate system of the high-resolution remote sensing image.

[0013] Overlay the vector boundary layer of supplementary cultivated land on the high-resolution remote sensing image to determine the spatial location and scope of each supplementary cultivated land patch;

[0014] Find no less than 5 identification points with geographic coordinates around each supplementary cultivated land patch, and crop the image according to the location of each supplementary cultivated land patch to ensure that the cropped image only contains the cultivated land patch area and the identification points around it.

[0015] Optionally, collecting drone images of the supplementary cultivated land area specifically includes:

[0016] Determine the geometric center coordinates of the supplementary cultivated land patch;

[0017] Control the drone to navigate to the geometric center of the cultivated land patch and collect drone images;

[0018] The drone imagery completely covers the boundaries of the supplementary cultivated land and ensures that the pixel identification points corresponding to no less than 5 selected geographic coordinates in the high-resolution remote sensing imagery can be identified in the drone imagery.

[0019] Optionally, the SIFT feature matching algorithm is used to extract feature points of the high-resolution remote sensing image and the UAV image, and the calculation and evaluation of the homography matrix specifically include the following steps:

[0020] Convert drone images and high-resolution remote sensing images into grayscale images and perform Gaussian blur denoising;

[0021] Detect feature points in UAV images and high-resolution remote sensing images based on the SIFT algorithm. Feature points include corners and edges.

[0022] Lowe's ratio test was used to screen matching points. Matching points are pairs of points with similar features found in the UAV image and the high-resolution remote sensing image using the SIFT algorithm, representing corresponding locations in the image.

[0023] By calculating the distance ratio between the closest matching point and the second closest matching point of each feature point, the calculation formula is: ,in is the distance between the feature point and the closest matching point, is the distance between the feature point and the second closest matching point. If the ratio is lower than the set threshold, the matching point pair is retained;

[0024] The FLANN algorithm is used to match feature points in UAV images and high-resolution remote sensing images by calculating the Euclidean distance between the descriptors of feature points.

[0025] Iteratively optimize the matching point pairs based on the RANSAC algorithm to eliminate incorrect matching points;

[0026] Convert the pixel coordinates of the matching points of the high-resolution remote sensing image into geographic coordinates;

[0027] The pixel coordinates of the matching points in the UAV image and the geographic coordinates of the matching points in the high-resolution remote sensing image are used to fit the homography transformation matrix between the images using the least squares method;

[0028] The reprojection error of the homography matrix is ​​calculated using the landmark points with known geographic coordinates in the UAV image. If the root mean square error of the reprojection error of all landmark points is greater than 10 pixels, the matching points are optimized and the homography matrix is ​​updated.

[0029] Optionally, optimize the matching points and update the homography matrix. Specifically, adjust the threshold of Lowe's ratio test, optimize the neighbor search parameters of FLANN algorithm, and improve the false matching rejection strategy of RANSAC algorithm to optimize SIFT matching points and obtain the optimal homography matrix.

[0030] Optionally, a homography matrix is ​​used to perform perspective transformation to map the supplementary cultivated land patches of the high-resolution remote sensing image to the drone image. Specifically, the following steps are performed:

[0031] Based on the homography matrix, the boundary points of the supplementary cultivated land patches in the high-resolution remote sensing image are transformed point by point, and the coordinates of the vector boundary layer are mapped to the UAV image coordinate system to achieve accurate alignment between the geospatial coordinates and the image pixel coordinates.

[0032] Optionally, segmenting the cultivated land and non-cultivated land areas within the supplementary cultivated land patch based on the YOLO algorithm and calculating their geographical areas and proportions respectively through the homography matrix specifically includes the following steps:

[0033] After superimposing the supplementary cultivated land patches on the drone image, the drone image is semantically segmented into cultivated land and non-cultivated land within the patch range, and the pixel-level segmentation results are output;

[0034] The polygonal image coordinates of the segmented cultivated land and non-cultivated land are converted into geographic coordinates through the inverse matrix of the homography matrix, and the actual area of ​​the cultivated land and non-cultivated land of the supplementary plot is calculated;

[0035] Calculate the area ratio of cultivated land to non-cultivated land in the geographic coordinate system and generate a restoration progress assessment report for supplementary cultivated land.

[0036] According to another aspect of the present application, a supplementary cultivated land restoration assessment system is provided, comprising:

[0037] The multi-source data acquisition module is used to access high-resolution remote sensing image data and overlay the cultivated land area; it also plans the UAV flight path and triggers image acquisition commands to ensure overlap with the high-resolution remote sensing image;

[0038] The image registration module performs denoising and grayscale normalization operations, unifies the imaging conditions of high-resolution remote sensing images and UAV images, generates descriptors through SIFT multi-scale feature detection, combines FLANN matching with RANSAC optimization to calculate the homography matrix, and loads the homography matrix to project the supplementary cultivated land patches in the high-resolution remote sensing image onto the UAV image.

[0039] The cultivated land segmentation module uses the YOLO algorithm to segment cultivated land and non-cultivated land pixel areas within the cultivated land patch, calculates the cultivated land coverage rate and non-cultivated land ratio, and outputs quantitative indicators;

[0040] The assessment report generation module is used to generate a supplementary arable land assessment report with maps.

[0041] According to another aspect of the present application, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned supplementary cultivated land restoration assessment method are implemented.

[0042] The beneficial effects of the supplementary cultivated land restoration assessment method, system and medium provided in this application are:

[0043] Leveraging the flexibility and short acquisition cycles of drones, this approach achieves precise geographic coordinate mapping of multi-source data. Specifically, the SIFT feature matching algorithm is applied to drone imagery and high-resolution remote sensing imagery, enabling pixel-to-pixel coordinate conversion of drone imagery. The supplementary cultivated land patches are then mapped onto the drone imagery, ensuring the accuracy of cultivated land area verification. Furthermore, the YOLO algorithm is used to perform instance segmentation on the cultivated land patches mapped by drone imagery, accurately distinguishing between cultivated and non-cultivated land within the geographic patches, providing solid data support for quantitative assessment of cultivated land area. This solution, through a closed technical loop of "precise coordinate mapping - intelligent image analysis - dynamic monitoring and early warning," establishes a comprehensive solution covering data collection, processing, and decision support. This effectively addresses long-standing challenges in supplementary cultivated land supervision, such as time lags, insufficient accuracy, and high costs, and provides reliable technical support for intelligent monitoring of supplementary cultivated land. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0045] in:

[0046] Figure 1 This is a flow chart of a method for evaluating the restoration of supplementary cultivated land, as shown in one embodiment of the present application;

[0047] Figure 2 It is a structural diagram of a supplementary cultivated land restoration assessment system shown in one embodiment of the present application. DETAILED DESCRIPTION

[0048] To facilitate understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings. The accompanying drawings illustrate preferred embodiments of the present application. However, the present application may be implemented in many other forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the disclosure of the present application.

[0049] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application pertains. The terms used herein in the specification of this application are for the purpose of describing specific embodiments only and are not intended to limit this application.

[0051] According to one aspect of the present application, an embodiment of the present application provides a method for evaluating the restoration of supplementary cultivated land, such as Figure 1 As shown, the supplementary cultivated land restoration assessment method includes the following steps:

[0052] S1. Obtain vector boundary layers and high-resolution remote sensing images of the supplementary cultivated land areas.

[0053] Specifically, step S1 includes the following steps:

[0054] Obtain the vector boundary layer of the supplementary cultivated land from the cultivated land business system and obtain a high-resolution remote sensing image with a resolution better than 1 meter. The high-resolution remote sensing image should cover the entire vector boundary layer of the supplementary cultivated land. If the coordinates of the vector boundary layer of the supplementary cultivated land are inconsistent with those of the high-resolution remote sensing image, use the seven-parameter transformation method to transform the coordinates of the vector boundary layer to make them consistent with the coordinate system of the remote sensing image (CGCS2000).

[0055] Overlay the vector boundary layer of supplementary cultivated land on the remote sensing image to determine the spatial location and scope of each supplementary cultivated land patch;

[0056] Find no less than 5 geographic coordinate identification points around each supplementary cultivated land patch, crop the image according to the location of each supplementary cultivated land patch, and ensure that the cropped image only contains the map patch area and its surrounding geographic identification points.

[0057] S2. Deploy the drone and navigate to the geometric center coordinates of the supplementary cultivated land patch to collect drone images of the supplementary cultivated land area.

[0058] Specifically, step S2 includes the following steps:

[0059] Determine the geometric center coordinates of the supplementary cultivated land patch;

[0060] Control the drone to navigate to the geometric center of the cultivated land patch and collect drone images;

[0061] The drone image should completely cover the boundary of the supplementary cultivated land and ensure that the pixel identification points corresponding to no less than 5 selected geographic coordinates in the remote sensing image can be identified in the drone image.

[0062] S3. Use the SIFT feature matching algorithm to extract feature points from high-resolution remote sensing images and UAV images, and calculate and evaluate the homography matrix.

[0063] Specifically, step S3 includes the following steps:

[0064] The UAV images and high-resolution remote sensing images were converted into grayscale images and denoised using Gaussian blurring with a 5×5 Gaussian kernel;

[0065] Detect significant feature points such as corners and edges in UAV images and high-resolution remote sensing images based on the SIFT algorithm;

[0066] Lowe's ratio test was used to screen matching points. Matching points are pairs of points with similar features found in UAV images and high-resolution remote sensing images using the SIFT algorithm, representing corresponding locations in the images.

[0067] By calculating the distance ratio between the closest matching point and the second closest matching point of each feature point, the calculation formula is: ,in is the distance between the feature point and the closest matching point, is the distance between the feature point and the second closest matching point. If the ratio is lower than the set threshold, the matching point pair (i.e. the closest matching point and the second closest matching point) is retained. It should be noted that the threshold should be set below 0.7 to balance accuracy and the number of matching points and effectively eliminate false matches.

[0068] Use the FLANN algorithm to match feature points in two images by calculating the Euclidean distance between feature descriptors;

[0069] Iteratively optimize the matching point pairs based on the RANSAC algorithm to eliminate incorrect matching points;

[0070] Convert the pixel coordinates of the matching points of the high-resolution remote sensing image into geographic coordinates;

[0071] Using the pixel coordinates of the matching points in the drone image and the geographic coordinates of the matching points in the high-resolution remote sensing image, the homography transformation matrix between the images is fitted using the least squares method. Specifically, the homography matrix H can be calculated using the following relationship:

[0072]

[0073] Where (x, y) is the geographic coordinate of the remote sensing image, (x', y') is the pixel coordinate of the drone image, w' is the weight factor, and H is the desired homography matrix;

[0074] Using landmarks with known geographic coordinates in drone imagery, the reprojection error of the homography matrix is ​​calculated. If the root mean square error of the reprojection error for all landmarks is greater than 10 pixels, the matching points are optimized and the homography matrix is ​​updated. This optimization process optimizes SIFT matching points by adjusting the Lowe's ratio test threshold, optimizing the nearest neighbor search parameters of the FLANN algorithm, and improving the false match rejection strategy of the RANSAC algorithm to obtain the optimal homography matrix.

[0075] S4. Use the homography matrix to perform perspective transformation and map the supplementary cultivated land patches of the high-resolution remote sensing imagery into the UAV imagery;

[0076] Specifically, step S4 includes: performing perspective transformation on the boundary points of the supplementary cultivated land patches in the high-resolution remote sensing image point by point based on the homography matrix H, mapping the vector boundary layer coordinates (x, y) to the UAV image coordinate system (x', y'), and realizing the precise registration of the geospatial coordinates and the image pixel coordinates.

[0077] S5. Segment the cultivated land and non-cultivated land within the supplementary cultivated land patch using the YOLOv11 algorithm, and calculate their respective geographic areas and proportions using the homography matrix;

[0078] Specifically, step S5 includes the following steps: after superimposing the supplementary cultivated land patches on the UAV image, semantically segmenting the cultivated land and non-cultivated land in the UAV image within the patch range, and outputting the pixel-level segmentation results;

[0079] The polygonal image coordinates of the segmented cultivated land and non-cultivated land are converted into geographic coordinates using the inverse matrix of the homography matrix H, and the actual area of ​​the cultivated land and non-cultivated land in the supplementary plot is calculated. The formula for converting a point (x', y') in the image coordinate system to a point (x, y) in the geographic coordinate system is as follows:

[0080]

[0081] Calculate the area ratio of cultivated land to non-cultivated land in the geographic coordinate system and generate a restoration progress assessment report for supplementary cultivated land.

[0082] The supplementary cultivated land restoration assessment method in the embodiment of the present application takes advantage of the flexibility and short acquisition cycle of drones to achieve precise mapping of geographic coordinates of multi-source data collaboration. Specifically, the SIFT feature matching algorithm is applied to drone images and high-resolution remote sensing images, which solves the geographic conversion of pixel coordinates of drone images and maps the supplementary cultivated land patches onto drone images to ensure the accuracy of cultivated land scope verification. At the same time, the YOLO algorithm is used to perform instance segmentation on the cultivated land patch range mapped by the drone image, accurately distinguishing cultivated land from non-cultivated land areas within the geographic patch, and providing solid data support for the quantitative assessment of cultivated land area. Through the technical closed loop of "precise coordinate mapping-intelligent image analysis-dynamic monitoring and early warning", this solution has built a full-chain solution covering data collection, processing, and decision support, effectively solving the long-term pain points of time lag, insufficient accuracy, and high cost in the supervision of supplementary cultivated land, and providing reliable technical support for the intelligent monitoring of supplementary cultivated land.

[0083] According to another aspect of the present application, the embodiment of the present application also provides a supplementary farmland restoration assessment system, such as Figure 2 Shown, including:

[0084] The multi-source data acquisition module is used to access high-resolution remote sensing image data and overlay the cultivated land area; it also plans the UAV flight path and triggers image acquisition commands to ensure overlap with the high-resolution remote sensing image;

[0085] The image registration module performs denoising and grayscale normalization operations, unifies the imaging conditions of high-resolution remote sensing images and UAV images, generates descriptors through SIFT multi-scale feature detection, combines FLANN matching with RANSAC optimization to calculate the homography matrix, and loads the homography matrix to project the supplementary cultivated land patches in the high-resolution remote sensing image onto the UAV image.

[0086] The cultivated land segmentation module uses the YOLO algorithm to segment cultivated land and non-cultivated land pixel areas within the cultivated land patch, calculates the cultivated land coverage rate and non-cultivated land ratio, and outputs quantitative indicators;

[0087] The assessment report generation module is used to generate a supplementary arable land assessment report with maps.

[0088] In addition, an embodiment of the present application also discloses a computer-readable storage medium.

[0089] Specifically, the computer-readable medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the supplementary arable land restoration assessment method.

[0090] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] The above embodiments merely illustrate several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for evaluating the restoration of supplementary cultivated land, characterized in that: The following steps are involved: Obtain vector boundary layers and high-resolution remote sensing images of supplementary cultivated land; Deploy the drone and navigate to the geometric center coordinates of each supplementary cultivated land patch to collect drone images of the supplementary cultivated land area; Use the SIFT feature matching algorithm to extract feature points from high-resolution remote sensing images and UAV images, and calculate and evaluate the homography matrix; Use homography matrix to perform perspective transformation to map the supplementary cultivated land patches of high-resolution remote sensing imagery into UAV imagery; The cultivated land and non-cultivated land areas within the supplementary cultivated land map are segmented based on the YOLO algorithm, and their geographical areas and proportions are calculated using the homography matrix. The method of extracting feature points of high-resolution remote sensing images and drone images using the SIFT feature matching algorithm and calculating and evaluating the homography matrix specifically includes the following steps: Convert drone images and high-resolution remote sensing images into grayscale images and perform Gaussian blur denoising; Detect feature points in drone images and high-resolution remote sensing images based on the SIFT algorithm, including corner points and edges; Lowe's ratio test is used to screen matching points. Matching points are pairs of points with similar features found in the UAV image and high-resolution remote sensing image using the SIFT algorithm, representing corresponding locations in the image. By calculating the distance ratio between the closest matching point and the second closest matching point of each feature point, the calculation formula is: Where d1 is the distance between the feature point and the closest matching point, and d2 is the distance between the feature point and the second closest matching point. If the ratio is lower than the set threshold, the matching point pair is retained; The FLANN algorithm is used to match feature points in UAV images and high-resolution remote sensing images by calculating the Euclidean distance between the descriptors of feature points. Iteratively optimize the matching point pairs based on the RANSAC algorithm to eliminate incorrect matching points; Convert the pixel coordinates of the matching points of the high-resolution remote sensing image into geographic coordinates; The pixel coordinates of the matching points in the UAV image and the geographic coordinates of the matching points in the high-resolution remote sensing image are used to fit the homography transformation matrix between the images using the least squares method; The reprojection error of the homography matrix is ​​calculated using the landmark points with known geographic coordinates in the UAV image. If the root mean square error of the reprojection error of all landmark points is greater than 10 pixels, the matching points are optimized and the homography matrix is ​​updated.

2. The method for evaluating the restoration of supplementary cultivated land according to claim 1, characterized in that: The acquisition of the vector boundary layer and the high-resolution remote sensing image of the supplementary cultivated land specifically includes the following steps: Obtain a vector boundary layer for the supplementary cultivated land from the cultivated land business system, and obtain a high-resolution remote sensing image with a resolution better than 1 meter, where the size of the high-resolution remote sensing image covers the entire vector boundary layer for the supplementary cultivated land; if the coordinates of the vector boundary layer for the supplementary cultivated land are inconsistent with those of the high-resolution remote sensing image, perform a coordinate conversion on the vector boundary layer to make it consistent with the coordinate system of the high-resolution remote sensing image; Overlay the vector boundary layer of supplementary cultivated land on the high-resolution remote sensing image to determine the spatial location and scope of each supplementary cultivated land patch; Find no less than 5 identification points with geographic coordinates around each supplementary cultivated land patch, and crop the image according to the location of each supplementary cultivated land patch to ensure that the cropped image only contains the cultivated land patch area and the identification points around it.

3. The method for evaluating the restoration of supplementary cultivated land according to claim 2, characterized in that: The acquisition of drone images of the supplementary cultivated land area specifically includes: Determine the geometric center coordinates of the supplementary cultivated land patch; Control the drone to navigate to the geometric center of the cultivated land patch and collect drone images; The drone image completely covers the boundary of the supplementary cultivated land and ensures that the pixel identification points corresponding to no less than 5 geographic coordinates selected in the high-resolution remote sensing image can be identified in the drone image.

4. The method for evaluating the restoration of supplementary cultivated land according to claim 1, wherein: The optimization of matching points and updating of the homography matrix specifically optimizes SIFT matching points by adjusting the threshold of Lowe's ratio test, optimizing the neighbor search parameters of the FLANN algorithm, and improving the false matching elimination strategy of the RANSAC algorithm to obtain the optimal homography matrix.

5. The method for evaluating the restoration of supplementary cultivated land according to claim 1, wherein: The use of the homography matrix to perform perspective transformation to map the supplementary cultivated land patches of the high-resolution remote sensing image to the drone image specifically includes: Based on the homography matrix, the boundary points of the supplementary cultivated land patches in the high-resolution remote sensing image are transformed point by point, and the coordinates of the vector boundary layer are mapped to the UAV image coordinate system to achieve accurate alignment between the geospatial coordinates and the image pixel coordinates.

6. The method for evaluating the restoration of supplementary cultivated land according to claim 1, characterized in that: The method of segmenting the cultivated land and non-cultivated land areas within the supplementary cultivated land map based on the YOLO algorithm and calculating their geographical areas and proportions respectively through the homography matrix specifically includes the following steps: After superimposing the supplementary cultivated land patches on the drone image, the drone image is semantically segmented into cultivated land and non-cultivated land within the patch range, and the pixel-level segmentation results are output; The polygonal image coordinates of the segmented cultivated land and non-cultivated land are converted into geographic coordinates through the inverse matrix of the homography matrix, and the actual area of ​​the cultivated land and non-cultivated land of the supplementary plot is calculated; Calculate the area ratio of cultivated land to non-cultivated land in the geographic coordinate system and generate a restoration progress assessment report for supplementary cultivated land.

7. A supplementary cultivated land restoration assessment system, characterized in that: include: Multi-source data acquisition module, used to access high-resolution remote sensing image data and overlay and supplement the scope of cultivated land areas; and planning the drone’s flight path, triggering image acquisition commands, and ensuring overlap with high-resolution remote sensing imagery; The image registration module performs denoising and grayscale normalization operations, unifies the imaging conditions of high-resolution remote sensing images and drone images, generates descriptors through SIFT multi-scale feature detection, combines FLANN matching with RANSAC optimization to calculate the homography matrix, and loads the homography matrix to project the supplementary cultivated land patches in the high-resolution remote sensing image onto the drone image. Calculating the homography matrix specifically includes the following steps: Convert drone images and high-resolution remote sensing images into grayscale images and perform Gaussian blur denoising; Detect feature points in drone images and high-resolution remote sensing images based on the SIFT algorithm, including corner points and edges; Lowe's ratio test is used to screen matching points. Matching points are pairs of points with similar features found in the UAV image and high-resolution remote sensing image using the SIFT algorithm, representing corresponding locations in the image. By calculating the distance ratio between the closest matching point and the second closest matching point of each feature point, the calculation formula is: Where d1 is the distance between the feature point and the closest matching point, and d2 is the distance between the feature point and the second closest matching point. If the ratio is lower than the set threshold, the matching point pair is retained; The FLANN algorithm is used to match feature points in UAV images and high-resolution remote sensing images by calculating the Euclidean distance between the descriptors of feature points. Iteratively optimize the matching point pairs based on the RANSAC algorithm to eliminate incorrect matching points; Convert the pixel coordinates of the matching points of the high-resolution remote sensing image into geographic coordinates; The pixel coordinates of the matching points in the UAV image and the geographic coordinates of the matching points in the high-resolution remote sensing image are used to fit the homography transformation matrix between the images using the least squares method; The reprojection error of the homography matrix is ​​calculated using the landmarks with known geographic coordinates in the UAV image. If the root mean square error of the reprojection error of all landmarks is greater than 10 pixels, the matching points are optimized and the homography matrix is ​​updated. The cultivated land segmentation module uses the YOLO algorithm to segment cultivated land and non-cultivated land pixel areas within the cultivated land patch, calculates the cultivated land coverage rate and non-cultivated land ratio, and outputs quantitative indicators; The assessment report generation module is used to generate a supplementary arable land assessment report with maps.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the supplementary cultivated land restoration assessment method according to any one of claims 1 to 6 are implemented.

Images