A farmland construction supervision method and system and a medium

Abstract

The application discloses a farmland construction supervision method and system and a medium, and relates to the technical field of farmland construction supervision. The method comprises the following steps: collecting high-resolution remote sensing images of a farmland project area at key time nodes, inputting the pre-trained ground feature classification model for pixel-by-pixel reasoning, and obtaining a ground feature classification mask; performing vectorization processing on the ground feature classification mask to obtain ground feature attribute vector data; comparing and analyzing the ground feature attribute vector data and construction planning vector data to obtain farmland construction progress; performing spatial overlay analysis on the ground feature attribute vector data at different key time nodes, extracting ground feature type change polygons and corresponding change information, comparing and analyzing the ground feature type change polygons and the corresponding change information with farmland construction specifications, and identifying illegal construction behaviors. The application realizes full-cycle fine supervision of farmland construction progress quantification and automatic identification of illegal behaviors.

Description

Technical Field

[0001] This invention relates to the field of farmland construction supervision technology, specifically to a farmland construction supervision method, system, and medium. Background Technology

[0002] With the development of satellite remote sensing technology, using remote sensing imagery for farmland monitoring has gradually become a research hotspot. Existing technologies already include relevant schemes for farmland monitoring using remote sensing imagery.

[0003] For example, CN120495909A discloses a remote sensing monitoring method and system for high-standard farmland use patterns through spatial-spectral fusion. This method involves multi-temporal synthesis and preprocessing of remote sensing images, calculating the Normalized Difference Vegetation Index (NDVI), analyzing phenological characteristics, and constructing a farmland use map based on prior knowledge of farmland use to monitor farmland use patterns. CN117557414A discloses a farmland monitoring method based on automatic interpretation of remote sensing images. This method uses a farmland interpretation model to automatically interpret current remote sensing images to obtain binary mask images. After extracting farmland patches, it compares these patches with those from the previous period to obtain farmland change information.

[0004] However, all the aforementioned existing technical solutions share a common drawback. Specifically, the binary masking method can only distinguish between arable land and non-arable land, outputting only a binary result of "whether arable land has changed," failing to answer key regulatory questions such as "what type of change" and "whether the change is compliant." In other words, while this solution can detect changes, it cannot identify what those changes are. In farmland construction scenarios, a piece of land changing from unused land to arable land is considered compliant construction; changing from arable land to construction land is considered illegal occupation. Both are represented as "arable land / non-arable land" changes under the binary mask, making effective differentiation impossible. Similarly, relying on a single NDVI value for feature identification provides extremely limited feature information, unable to support refined identification of multiple types of features such as field roads, ponds, and construction land, and also unable to correlate feature changes with specific construction progress or violation types.

[0005] Therefore, there is an urgent need to propose a method that can deeply integrate remote sensing image change detection results with farmland construction supervision, so as to achieve accurate and automated supervision of the entire farmland construction process by finely identifying specific changes in land cover types. Summary of the Invention

[0006] The purpose of this invention is to provide a method, system, and medium for monitoring farmland construction, in order to solve the technical problem that existing farmland construction monitoring technologies can only detect whether there are changes, but cannot effectively correlate information on changes in land features with construction progress and violations, and thus achieve refined monitoring throughout the entire process.

[0007] This invention is achieved through the following technical solution:

[0008] In a first aspect, the present invention provides a method for supervising farmland construction, comprising:

[0009] High-resolution remote sensing images of the farmland project area at key time points are collected, including each preset monitoring point during the farmland construction period.

[0010] The high-resolution remote sensing image is input into a pre-trained land cover classification model for pixel-by-pixel inference to obtain the land cover classification mask of the farmland project area at the corresponding key time nodes.

[0011] The land feature classification mask is vectorized to obtain land feature attribute vector data;

[0012] The farmland construction progress is obtained by comparing and analyzing the vector data of the land feature attributes with the vector data of the construction plan.

[0013] Spatial overlay analysis of land cover attribute vector data at different key time points was performed to extract land cover type change patches and corresponding change information;

[0014] By comparing and analyzing the changed patches and change information with farmland construction standards, illegal construction activities during the construction process can be identified.

[0015] In some implementations, the key time points also include a post-completion verification point; the method further includes:

[0016] Spatial overlay analysis is performed on the vector data of land cover attributes between different verification nodes to extract land cover type change patches and corresponding change information; and / or, spatial overlay analysis is performed on the vector data of land cover attributes of verification nodes and the vector data of land cover attributes of preset monitoring nodes to extract land cover type change patches and corresponding change information.

[0017] By comparing and analyzing the changes in land cover types and corresponding change information with the farmland project planning scheme, illegal land use after completion can be identified.

[0018] In some implementations, the preset monitoring nodes include a commencement time node, a completion and acceptance node, and at least one phased acceptance node during the construction process.

[0019] In some embodiments, the method further includes, before inputting the high-resolution remote sensing image into a pre-trained land cover classification model for pixel-by-pixel inference:

[0020] The high-resolution remote sensing image is subjected to standardized preprocessing; the standardized preprocessing includes at least one of radiometric calibration, atmospheric correction and image registration.

[0021] In some implementations, the land feature classification mask is vectorized to obtain land feature attribute vector data, including:

[0022] In the land feature classification mask, pixels belonging to the same land feature type and spatially connected are aggregated into independent vector plots;

[0023] Extract the spatial location and vector boundary of each vector vector patch, calculate the geometric area of ​​each vector vector patch, and assign the corresponding land feature type to obtain the land feature attribute vector data.

[0024] In some implementations, the change information of the land feature type change patch includes the land feature type before and after the change, as well as the change in area.

[0025] In some implementations, the changed patches and change information are compared and analyzed with farmland construction standards to identify illegal construction activities during the construction process, including:

[0026] Based on the land cover types before and after the change in the map patch, determine whether the change complies with farmland construction standards;

[0027] If it does not comply with regulations and lacks approval procedures, it will be determined as an illegal construction activity, and the vector boundary of the changed feature will be extracted as the illegal feature.

[0028] In some embodiments, the high-resolution remote sensing image is a sub-meter high-resolution optical remote sensing image.

[0029] In a second aspect, the present invention provides a farmland construction monitoring system for executing the farmland construction monitoring method according to any one of the first aspects of the present invention, the system comprising:

[0030] The image acquisition module is used to acquire high-resolution remote sensing images of the farmland project area at key time nodes, including each preset monitoring node during the farmland construction period.

[0031] The land cover classification module inputs the high-resolution remote sensing image into a pre-trained land cover classification model for pixel-by-pixel inference to obtain the land cover classification mask of the farmland project area at the corresponding key time nodes.

[0032] The vectorization module is used to vectorize the land feature classification mask to obtain land feature attribute vector data;

[0033] The feature extraction module is used to perform spatial overlay analysis on the vector data of land cover attributes at different key time points to extract feature type change features and corresponding change information.

[0034] The progress monitoring module is used to compare and analyze the vector data of the land feature attributes with the vector data of the construction plan to obtain the progress of farmland construction;

[0035] The violation monitoring module is used to compare and analyze the changed map patches and change information with farmland construction standards to identify illegal construction activities during the construction process.

[0036] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the farmland construction supervision method according to any one of the first aspects of the present invention.

[0037] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0038] In the above method of this invention, a pre-trained land cover classification model is used to perform pixel-by-pixel inference on multiple high-resolution remote sensing images to obtain refined land cover classification results at each time point; then, the regulatory objective is achieved through two independent comparison paths:

[0039] Path 1 involves spatially overlaying and comparing the vector data of land feature attributes with the vector data of construction planning, using whether the changed land feature types and spatial ranges conform to the planning as the standard to quantify the construction progress of various facilities;

[0040] Path 2 involves performing time-series overlay analysis on the vector data of land feature attributes at different time points, extracting the specific land feature types before and after the change of each changed patch, and then comparing them with farmland construction standards, thereby accurately mapping the "change in land feature type" to the "specific type of violation".

[0041] This invention bridges the semantic gap between binary change detection and regulatory business needs by acquiring information on the types of land features and changes before and after changes in changed patches. It enables remote sensing change detection results to directly serve core regulatory issues such as "how is the construction progress" and "what are the violations", thus achieving refined supervision throughout the entire process. Attached Figure Description

[0042] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0043] Figure 1 This is a flowchart of a farmland construction supervision method according to an embodiment of the present invention;

[0044] Figure 2 This is a flowchart of a method for vectorizing a land cover classification mask according to an embodiment of the present invention;

[0045] Figure 3 This is an application flowchart of a farmland construction supervision method according to an embodiment of the present invention;

[0046] Figure 4 This is a block diagram of a farmland construction monitoring system according to an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0048] This invention provides a method, system, and medium for monitoring farmland construction, aiming to address the core shortcomings of existing high-standard farmland remote sensing monitoring technologies. These shortcomings include limited monitoring features, the inability to classify farmland into arable / non-arable land, the inability to analyze specific changes, difficulty in identifying construction progress, and the inability to accurately determine violations. This invention overcomes the limitations of existing technologies in achieving comprehensive, refined, and precise dynamic monitoring of high-standard farmland, thereby improving the intelligent monitoring level of high-standard farmland construction and management. It is applicable to various farmland construction projects, including high-standard farmland construction, for progress monitoring and violation identification scenarios.

[0049] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0050] Please see Figure 1 , Figure 1 A flowchart of a farmland construction supervision method according to the present invention is shown, including the following steps:

[0051] S101, collects high-resolution remote sensing images of the farmland project area at key time points.

[0052] In this step, after identifying the target farmland project area, basic project data is entered based on the project construction information, such as vector boundaries of the project planning area, design and construction content, approved project schedule, and responsible entity, in both vector and text format. Key time nodes for project supervision are then defined based on the project construction information; these key time nodes may include pre-set supervision points during farmland construction.

[0053] In this embodiment, the preset monitoring nodes may include the commencement time node (i.e., the time node before formally entering construction, which can be determined as the latest image acquisition time before the formal construction of the project), the completion and acceptance node (i.e. the completion and acceptance time node agreed in the project contract, which can be determined as the latest image acquisition time after the completion date), and at least one phased acceptance node in the construction process (i.e., the image acquisition time corresponding to each phased acceptance within the project construction cycle).

[0054] For each key time point, acquire high-resolution optical remote sensing images of the same farmland project area that are cloudless, have minimal fog, and possess acceptable imaging quality. Preferably, use sub-meter resolution optical remote sensing images with a spatial resolution better than 1 meter to ensure clear identification of details such as roads and ditches within the fields.

[0055] S102, high-resolution remote sensing images are input into a pre-trained land cover classification model for pixel-by-pixel inference to obtain the land cover classification mask of the farmland project area at the corresponding key time nodes.

[0056] In this embodiment, before inputting the images into the land cover classification model, the acquired images can undergo standardization preprocessing to eliminate spatiotemporal differences between images acquired at different times. Specifically, standardization preprocessing includes at least one of radiometric calibration, atmospheric correction, and image registration. Radiometric calibration converts the original grayscale values ​​of the images into physically meaningful radiance or reflectance values; atmospheric correction eliminates the influence of atmospheric scattering and absorption on surface reflectance information; and image registration ensures accurate spatial alignment of images from different periods.

[0057] Preferably, during image registration, the boundary of the farmland project planning vector is used as a reference. The size of the circumscribed rectangle is controlled by the distance between neighboring project vectors to obtain the circumscribed rectangle of the farmland project cluster area. After segmentation, remote sensing image data for each time node are obtained. The high-resolution remote sensing image after standardization preprocessing is input into a pre-trained land cover classification model for pixel-by-pixel inference. Standardization preprocessing ensures spatial consistency between multiple image periods, providing a reliable data foundation for subsequent analysis.

[0058] In this step, the land cover classification model adopts a semantic segmentation model based on deep learning. Its input is a high-resolution remote sensing image, and its output is a land cover classification mask of the same size as the input image. The value of each pixel in the mask represents the land cover category to which that pixel belongs. A full-feature model is preferred, supporting refined classification of various land cover types such as cultivated land, field roads, ponds and water surfaces, orchards, construction land, unused land, illegally occupied land, and background, avoiding classification bias caused by single features.

[0059] In some implementations, supervised training is used to train the land cover classification model. The specific steps are as follows:

[0060] First, obtain high-resolution remote sensing images (sub-meter level) of the sample area and the corresponding land cover classification labels.

[0061] The land cover classification labels cover a wide range of land cover information, including cultivated land, field roads, ponds and water surfaces, orchards, construction land, unused land, illegally occupied land, and background. Sub-meter resolution remote sensing images covering different regions, seasons, and lighting conditions are collected to construct training samples for land cover classification. Professional remote sensing interpreters annotate the land cover features in each image pixel-by-pixel, generating corresponding land cover classification label maps. The annotated image-label pairs are then divided into training, validation, and test sets according to a preset ratio (e.g., 8:1:1).

[0062] Then, using sub-meter high-resolution remote sensing images as input and corresponding land cover classification labels as supervision signals, the initial semantic segmentation model is trained under supervision.

[0063] The initial model during training can employ known semantic segmentation network architectures such as DeepLabV3+, U-Net, or Swin Transformer. Remote sensing imagery is input into the initial model to obtain the probability of each pixel corresponding to the land cover type predicted by the model output, thus obtaining the prediction mask. The loss between the prediction mask and the ground truth label (such as cross-entropy loss) is calculated. The model parameters are adjusted using the backpropagation algorithm, and training is iteratively continued until the model converges on the validation set.

[0064] Finally, the classification accuracy of the model is evaluated on the test set. Once the accuracy requirements are met, it can be used for actual inference.

[0065] When applied, the trained land cover classification model is loaded, and pixel-by-pixel inference is performed on the standardized preprocessed high-resolution remote sensing images of each regulatory node (construction start time node, various stage acceptance nodes, completion acceptance nodes, etc.), outputting the full-element land cover classification mask corresponding to the node. Each pixel value in the mask represents the land cover category at that location, such as cultivated land, field roads, ponds and water surfaces, orchards, construction land, unused land, etc.

[0066] It should be noted that in actual monitoring projects, image data at each time point is continuously acquired and analyzed in real time as the project progresses, rather than waiting to collect all image data for the entire construction cycle before performing analysis all at once.

[0067] S103, vectorize the ground feature classification mask to obtain ground feature attribute vector data.

[0068] In this step, the land cover classification masks obtained in step S102 are vectorized and converted into land cover attribute vector data. This data includes vector information for each type of land parcel within the complete farmland area, such as boundary vectors, location, and area.

[0069] In one specific implementation, the method for converting a land cover classification mask into land cover attribute vector data is as follows:

[0070] (1) Aggregate spatially connected pixels belonging to the same land cover type in the land cover classification mask into independent vector polygons. For example, aggregate all pixels in the mask that are classified as "field roads" and are interconnected into one or more vector polygons, with each polygon corresponding to a field road polygon.

[0071] (2) Extract the spatial location and vector boundary of each vector polygon, that is, determine the geographic coordinates and geometric shape of each vector polygon.

[0072] (3) Calculate the geometric area of ​​each vector patch.

[0073] (4) Assign the corresponding land feature type attributes to each vector patch.

[0074] After the above processing, the vector data of land cover attributes corresponding to the land cover classification masks for each region are obtained. (Based on the construction start time node...) mask For example, vectorization yields vector data. , It contains multiple vector polygons, each with attributes such as spatial location, vector boundary, feature type, and geometric area. Similarly, the masks for each stage of acceptance and the final acceptance are vectorized to obtain the corresponding vector data.

[0075] S104 compares and analyzes the vector data of land feature attributes with the vector data of construction planning to obtain the progress of farmland construction.

[0076] Construction planning vector data includes vector boundaries and planning type information for various planned facilities in the project planning scheme, such as planned farmland areas, planned field roads, and planned ditches. For example, it includes information on phased acceptance milestones. The vector data of the land feature attributes of point 1 are compared and analyzed with the vector data of the construction plan of the same period in the project planning scheme, and the nodes are calculated. The overlap rate between the completed area and the total planned area in the as-built drawings is used to monitor the progress of farmland construction. Completion and acceptance milestones will be determined accordingly. The vector data of land features are spatially overlaid with the vector data of construction planning to calculate the spatial overlap rate between the actual coverage of each land feature type and the planned vector boundary after completion, in order to track changes in arable land.

[0077] For a specific type of planned facility, the formula for calculating its spatial overlap rate is: Spatial overlap rate = Area actually built and matching the type. The total planned facility area is multiplied by 100%. When the land feature type corresponding to the land feature classification result of a certain area is consistent with the planned facility type, and the spatial overlap rate between the actual coverage area and the planned vector boundary reaches a preset threshold, the facility corresponding to that area is determined to be completed. For planned cultivated land, the land feature attribute vector data of the completion acceptance node with the category "cultivated land" can be overlaid with the planned cultivated land vector boundary, and the proportion of the overlapping area to the total planned cultivated land area can be calculated to obtain the cultivated land completion rate. For planned field roads, the vector map of the completion acceptance node with the category "field road" can be overlaid with the planned road vector boundary, and the length overlap can be calculated to obtain the road completion rate.

[0078] The proportion of the area (or length) of each type of completed facility to the total area (or length) of the planned facilities is used as a quantitative indicator of project construction progress. This quantitative indicator can be used to determine whether the project has been completed as planned and whether there are any delays or other issues.

[0079] S105 involves spatial overlay analysis of vector data of land cover attributes at different key time points to extract land cover type change patches and corresponding change information.

[0080] In this step, by spatially overlaying and analyzing the vector data of land cover attributes from two time points, we can identify the land cover types that have changed. Specifically, the vector data from the construction commencement time point... Vector data of each stage of acceptance. Vector data of completion and acceptance milestones Perform pairwise or global overlay analysis to obtain Changes in farmland during the process.

[0081] Spatial overlay analysis employs GIS spatial analysis methods, spatially superimposing vector maps from different periods to detect whether the land cover types at the same spatial location have changed over time. When the land cover types at a certain spatial location are inconsistent between two periods, the changed map patch at that location is extracted, and its change information is recorded. This change information includes the spatial location of the changed map patch, the land cover types before and after the change, and the amount of area change. For example, for a certain area at the start of construction... The land feature type is "unused land", at the phased acceptance node. If the land feature type changes to "arable land", then the changed map patch is extracted, and its location, the type before the change was "unused land", the type after the change is "arable land", and the amount of area change are recorded.

[0082] S106 compares and analyzes the changed map patches and change information with farmland construction standards to identify illegal construction activities during the construction process.

[0083] In this step, based on the land cover types before and after the change of the changed patch extracted in step S105, it is determined whether the change complies with the requirements of farmland construction specifications and relevant laws and regulations. If it does not comply and there are no legal approval procedures, it is determined to be an illegal construction activity, and the vector boundary of the changed patch is extracted as the output of the illegal patch.

[0084] Illegal construction practices include construction not following the plan, construction errors exceeding permissible limits, and perfunctory completion acceptance. For example, if at the completion acceptance stage... The land feature classification result within the vector boundary of a certain planned facility is still the result before construction started. If the original land cover type is inconsistent with the planned facility type and the actual coverage area overlaps with the planned vector space by less than a preset threshold (e.g., 80%), and there is no reasonable construction delay certificate or approval procedure, then there is a violation of construction not in accordance with the plan during the construction process.

[0085] For each identified violation, the vector boundary of the violation patch is automatically extracted, generating standardized vector data of the violation patch. The data fields include the spatial location of the violation patch, vector boundary, actual violation area, and violation type, providing accurate spatial basis for subsequent problem rectification, responsibility determination, and law enforcement investigation.

[0086] Ultimately, remote sensing images, land feature classification results, construction progress data, and violation records throughout the entire project lifecycle will be uniformly archived to form a traceable and regulatory file for the entire farmland construction process, supporting project verification, auditing, and long-term management.

[0087] In some implementations, key time points also include post-completion verification points. Post-completion verification points refer to one or more verification time points after the project is completed and accepted, and during the later maintenance phase, used to monitor the utilization and maintenance status of farmland after its construction.

[0088] For the high-resolution remote sensing images collected at the verification nodes, the same processing method as that used at the aforementioned monitoring nodes is employed to extract land cover classification masks and land cover attribute vector data. Spatial overlay analysis is then performed on the land cover attribute vector data from different verification nodes to extract land cover type change patches and corresponding change information. By comparing and analyzing the change patches and corresponding change information with the farmland project planning scheme, illegal land use after completion is identified, such as land use not in accordance with the farmland project planning scheme or unauthorized changes to land use.

[0089] Furthermore, spatial overlay analysis is performed on the feature attribute vector data of the verification nodes and the feature attribute vector data of the preset supervision nodes (such as the completion and acceptance nodes) to extract feature type change patches and corresponding change information. Based on the feature type change patches and corresponding change information, a comparative analysis is conducted with the farmland project planning scheme to identify illegal land use behavior after completion. Simultaneously, for each identified illegal user behavior, the vector boundary of the illegal patch is automatically extracted to generate standardized illegal patch vector data.

[0090] In actual supervision, illegal land use includes, but is not limited to, the following situations:

[0091] (1) Illegal occupation of cultivated land for non-agricultural purposes: based on the completion and acceptance milestones. Based on the land cover classification results, if The land feature type of the node is cultivated land, while the node is under later management. If the land feature type is changed to construction land or non-agricultural land such as ponds or water surfaces, and there are no legal approval procedures, it is judged as illegal occupation of cultivated land for non-agricultural purposes.

[0092] (2) Illegal occupation of arable land for non-grain purposes: based on the completion and acceptance milestones. Based on the land cover classification results, if The land feature type of the node is cultivated land, while the node is under later management. If the land use type is changed to non-grain crop planting land such as gardens or greenhouses, and there are no legal approval procedures, it is judged as illegal occupation of arable land for non-grain crops.

[0093] (3) Illegal abandonment of cultivated land: based on the completion and acceptance milestones. Based on the land cover classification results, if The land feature type of the node is cultivated land, while the node is under later management. and If the land cover type of both nodes in the previous year was unused land (abandoned state) and there was no legal abandonment certificate, then it was determined to be illegal abandonment of cultivated land.

[0094] This implementation, by setting up verification nodes after completion and conducting spatial overlay analysis, can effectively identify various illegal land use behaviors that occur after farmland is built, and achieve long-term management and supervision of farmland, making up for the shortcomings of existing technologies that only focus on the construction stage and neglect the later management and maintenance.

[0095] In some implementations, the pre-set monitoring nodes specifically include commencement time nodes, completion and acceptance nodes, and at least one phased acceptance node during the construction process. By setting multiple key time nodes throughout the entire farmland construction process, a complete time sequence from pre-construction to post-completion is formed, providing a data foundation for phased dynamic monitoring.

[0096] Among them, the start time node This refers to the most recent image acquisition time before the project's formal construction. The land cover classification results at this node reflect the original surface conditions before project construction, serving as a benchmark for subsequent change monitoring. (Phase Acceptance Node) This refers to the image acquisition time corresponding to each stage of acceptance during the project construction cycle. Depending on the project scale and duration, one or more stage acceptance nodes can be set, such as the completion node for foundation construction, the completion node for the main structure, etc. By comparing the feature classification results at each stage acceptance node with the pre-construction baseline, construction progress can be tracked in real time, and problems during construction can be identified and corrected promptly. Final acceptance node. The latest image acquisition time after the project contractually agreed completion date is used to comprehensively verify whether the project has been completed according to plan. It is the core basis for quantifying construction progress and determining acceptance.

[0097] In some implementations, see Figure 2 As shown, the specific steps for vectorizing the land cover classification mask to obtain land cover attribute vector data are as follows:

[0098] S201 aggregates spatially connected pixels belonging to the same land cover type in the land cover classification mask into independent vector plots.

[0099] This process can employ binary image connected component analysis to extract the connected components for each type of land cover in the mask. Specifically, for a certain land cover category (such as "farmland"), the mask is scanned pixel by pixel starting from the top left corner, and adjacent (four-neighbor or eight-neighbor) pixels of the same category are marked as the same connected region. After the scan is completed, each connected region becomes an independent set of pixels.

[0100] S202, extract the vector boundary of the cell set for each connected region.

[0101] This step can use an edge tracing algorithm (such as Moore's Neighborhood Tracking) to extract the outer contour of the connected region and convert the contour point sequence into a vector polygon. For connected regions with holes (i.e., containing other land cover types), the inner contour also needs to be extracted.

[0102] S203, calculate the geometric properties of each vector plot.

[0103] For each extracted vector polygon, calculate its geometric area (for projected coordinate systems, calculate the polygon area directly; for geographic coordinate systems, a projection transformation is required before calculation) and spatial attributes such as the geometric center location.

[0104] S204: Assign corresponding land feature type attributes to each vector map patch and write the land feature category to which the map patch belongs into the attribute table.

[0105] After the above processing, the resulting vector data of land features contains multiple vector patches. The attribute table of each patch records information such as patch number, land feature type, geometric area, spatial location (e.g., center point coordinates), and vector boundary coordinate string. This vector data format is compatible with common GIS data formats (such as Shapefile), facilitating subsequent spatial overlay analysis and comparison with construction planning vector data.

[0106] In some implementations, the change information of land cover type change patches includes the land cover type before and after the change, as well as the amount of area change. Specifically, when performing spatial overlay analysis on land cover attribute vector data from two different time points, for any change patch, the following information is recorded:

[0107] (1) Spatial location of the changing patch: represented by the geometric center coordinates or the circumscribed rectangle of the changing patch.

[0108] (2) Type of land cover before the change: that is, the type of land cover corresponding to the map patch at an earlier time point.

[0109] (3) Changed land cover type: that is, the land cover category corresponding to the map patch at a later time point.

[0110] (4) Area change: The difference between the area of ​​the patch after the change and the area of ​​the patch before the change. If the area increases, the area change is positive; if the area decreases, the area change is negative.

[0111] (5) Vector boundary of the changing patch: the geometric shape vector data of the changing patch.

[0112] The aforementioned change information is stored in a structured data format (such as tables or GeoJSON) to facilitate subsequent identification and statistical analysis of violations. For example, regarding "arable land..." The change of the "construction land" map patch, whose type was "arable land" before the change and "construction land" after the change, and the area change is the area of ​​the map patch (usually a negative value, indicating a reduction in arable land). This information can be directly used to determine the violation of the conversion of arable land to non-agricultural use.

[0113] In some implementations, a mapping table is pre-established to correspond to changes in land cover categories and types of business violations. This mapping table can be pre-set according to relevant laws, regulations, and construction specifications for farmland construction, as shown in Table 1. During the process of identifying violations based on changed land cover features and change information, the corresponding violation type can be directly output by referring to the mapping table.

[0114] Table 1 Mapping Table of Land Feature Category Changes and Business Violation Types

[0115]

[0116] Specifically, for each changed map patch extracted in step S105, the combination of land cover types before and after the change is checked against the aforementioned mapping table. If they are, it is initially identified as a suspected violation. The suspected violation map patches are then further verified to check whether they have valid approval procedures. If valid approval procedures exist (such as construction land approval, facility agricultural land registration, etc.), it is not considered a violation; if no valid approval procedures exist, it is ultimately determined to be an illegal construction activity.

[0117] Finally, for the map features identified as violations, their vector boundaries are extracted as violation map features, and violation map feature vector data is generated according to a preset format. The attribute fields of the violation map feature vector data include at least: violation map feature number, spatial location, vector boundary coordinates, actual violation area, violation type (such as "non-agricultural use of cultivated land", "non-grain use of cultivated land", "illegal abandonment of cultivated land", etc.), judgment basis (type of land features before and after the change), and discovery time.

[0118] The output vector data of illegal patches can be directly imported into a GIS platform for visualization, or exported in a standardized format (such as Shapefile or GeoJSON) for use by other systems, providing accurate spatial data support for subsequent on-site verification, problem rectification and accountability.

[0119] The following example, using the supervision of a high-standard farmland construction project, illustrates the practical application and effectiveness of the supervision method of this invention. Figure 3 This is a flowchart of the implementation method. The specific implementation method is as follows:

[0120] 1.1 Project basic information entry and key time node delineation.

[0121] Enter basic data for high-standard farmland projects, including vector and text information such as the project planning scope, design and construction content, approved project schedule, and responsible entities.

[0122] According to the construction plan, three project supervision nodes and one post-completion verification node have been designated:

[0123] : Pre-construction benchmark node (the time of the most recent image acquisition before the formal construction of the project);

[0124] : Monitoring nodes during construction (the time of image acquisition for a certain period within the project's construction cycle);

[0125] : Completion verification milestone (the time of the most recent image collection after the completion date stipulated in the project contract);

[0126] : Post-construction maintenance milestones (time of image acquisition for verification after completion).

[0127] against At each time point, high-resolution optical remote sensing images of the same area with no clouds, little fog, and acceptable imaging quality were acquired to ensure consistent spatial resolution. The images underwent standardized preprocessing, including radiometric calibration, atmospheric correction, and registration, to ensure complete spatial alignment of images from multiple periods. Through image mosaicking and cropping, the size of the circumscribed rectangle was controlled to obtain remote sensing image data of each time point for the high-standard farmland project cluster area after the circumscribed rectangle was cut.

[0128] 1.2 Node-by-node feature extraction based on a full-element feature classification model.

[0129] The high-standard farmland full-element land cover classification model, which has been trained, is used to infer the remote sensing image data at various time points. The model supports fine classification, and the land cover types include: cultivated land, field roads, ponds and water surfaces, orchards, construction land, unused land, illegally occupied land, and background.

[0130] To each , , , After preprocessing, the image is used for pixel-by-pixel inference using a full-feature land cover classification model. The model outputs the land cover classification mask corresponding to the full-feature land cover classification of each node. After vectorizing the mask, the category information and feature area information are written into the attribute table to obtain a vector result containing the current land cover attributes and area. , , , .

[0131] 1.3 Analysis of vector attributes and spatial variation information of multi-node land cover classification.

[0132] Will , , , The ground feature vector corresponding to the node , , , Spatial overlay analysis was performed to extract land cover type change patches in the project area, and the spatial location, area, shape, and land cover type information before and after the change of the change patches were calculated.

[0133] 1.4 The quantitative assessment and acceptance of the construction progress of high-standard farmland were deemed to be mere formalities and thus violated regulations.

[0134] Construction monitoring nodes Completion milestones Full-element land cover classification results , Based on the core criteria, and combined with the vector data in the project planning and design scheme and construction standard requirements, the vector data of the as-built drawings is calculated. , Spatial overlap rate. Wherein, spatial overlap rate = area of ​​all feature classification results. Total planned area in the as-built drawings × 100%.

[0135] use The overlap rate between the construction area and the total planned area in the as-built drawings is used to assess the progress of high-standard farmland construction. The overlap rate between the construction area and the total planned area in the as-built drawings is used to verify and inspect the construction of high-standard farmland. The specific implementation process is as follows:

[0136] (1) Identification of existing facilities: Spatial overlay and comparison are performed between the node feature classification mask and the planned facility vector boundary. In the land feature classification results, if the land feature type in a certain area matches the planned facility type, and the actual coverage area overlaps with the planned vector space by ≥80%, it is determined to be an existing facility. Specifically, this includes:

[0137] A. Cultivated Land: Planning scope of paddy fields / dry land and If the overlap of nodal cultivated land (paddy field / dry land) classification patches is ≥90%, it is determined to be cultivated land that has been developed.

[0138] B. Farm roads: Planned road vector boundaries and If the length overlap of the "field road" category map patch is ≥90%, it is determined to be a constructed farm road;

[0139] (2) Acceptance inspection becomes a mere formality: when the planned facility's vector boundary is within... If the node's feature classification result is still the original feature type before planning, or inconsistent with the planned facility type, or if the actual coverage area overlaps with the planned vector space by less than 80%, and there is no reasonable proof of construction delay, it will be determined as a completed node. The acceptance process was merely a formality.

[0140] 1.5 Accurate identification and location of construction violations.

[0141] by The land cover classification results of the nodes are used as a reference for comparison. , , Changes in land cover types at each node are assessed in conjunction with planning requirements to determine whether they comply with high-standard farmland construction standards. Changes that do not comply with standards and lack legal approval are deemed violations. A system for tracking changes in land cover types is also established. The correspondence between "business violation types".

[0142] 1.6 Output regulatory results, violation maps, progress reports, and violation lists.

[0143] Embodiments of the present invention also provide a farmland construction monitoring system for executing the farmland construction monitoring method of any of the above embodiments of the present invention. Figure 4 As shown, the system includes:

[0144] The image acquisition module 201 is used to acquire high-resolution remote sensing images of the farmland project area at key time points.

[0145] The ground feature classification module 202 is used to load a pre-trained ground feature classification model, perform inference calculations on the input remote sensing images, and output the ground feature classification mask for the corresponding key time nodes.

[0146] Vectorization module 203 is used to vectorize the ground feature classification mask to obtain ground feature attribute vector data.

[0147] The feature extraction module 204 is used to perform spatial overlay analysis on the vector data of land cover attributes at different key time points to extract feature type change features and corresponding change information.

[0148] The progress monitoring module 205 is used to compare and analyze the vector data of land feature attributes with the vector data of construction planning to obtain the progress of farmland construction.

[0149] The violation monitoring module 206 is used to compare and analyze the changed map patches and change information with farmland construction standards to identify illegal construction activities during the construction process.

[0150] Key time nodes include pre-set supervision nodes during farmland construction (start-up time node, phased acceptance node, completion acceptance node, etc.) as well as verification nodes after completion.

[0151] The image acquisition module 201 can automatically acquire optical remote sensing images of a specified area, time, and resolution by connecting to the data interface of the remote sensing satellite data service platform, or it can acquire image data through the UAV aerial photography system.

[0152] Specifically, the vectorization module 203 aggregates spatially connected pixels belonging to the same land cover type in the land cover classification mask into independent vector patches, and extracts attribute information such as spatial location, vector boundary, geometric area, and land cover type. The patch extraction module 204 can specifically employ GIS spatial overlay analysis methods to detect changes in land cover types at the same spatial location at different times, generating changed patches and their change information. The progress monitoring module 205 quantifies the construction progress of various planned facilities by calculating the spatial overlap rate between the actual coverage of each land cover type after completion and the planned vector boundary. The violation monitoring module 206 automatically determines violations and outputs violation patches based on the land cover types before and after the changes in the changed patches, combined with the farmland construction standard mapping table and approval procedure information.

[0153] Embodiments of the present invention also provide an electronic device comprising a processor and a memory, wherein the number of processors may be one or more. The memory, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules. The processor executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory, thereby implementing the farmland construction monitoring method of any of the above embodiments of the present invention.

[0154] The memory may primarily comprise a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory may include high-speed random access memory (RAM) and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, the memory may further include memory remotely located relative to the processor, which can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0155] Embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the farmland construction supervision method of any embodiment of the present invention.

[0156] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0157] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0158] Embodiments of the present invention also provide a computer program product that, when run on a computer, causes the computer to execute the farmland construction supervision method of any of the above embodiments of the present invention.

[0159] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for supervising farmland construction, characterized in that, include: High-resolution remote sensing images of the farmland project area at key time points are collected, including each preset monitoring point during the farmland construction period. The high-resolution remote sensing image is input into a pre-trained land cover classification model for pixel-by-pixel inference to obtain the land cover classification mask of the farmland project area at the corresponding key time nodes. The land feature classification mask is vectorized to obtain land feature attribute vector data; The farmland construction progress is obtained by comparing and analyzing the vector data of the land feature attributes with the vector data of the construction plan. Spatial overlay analysis of land cover attribute vector data at different key time points was performed to extract land cover type change patches and corresponding change information; By comparing and analyzing the changed patches and change information with farmland construction standards, illegal construction activities during the construction process can be identified.

2. The farmland construction supervision method according to claim 1, characterized in that, The key time points also include post-completion verification points; the method further includes: Spatial overlay analysis is performed on the vector data of land cover attributes between different verification nodes to extract land cover type change patches and corresponding change information; and / or, spatial overlay analysis is performed on the vector data of land cover attributes of verification nodes and the vector data of land cover attributes of preset monitoring nodes to extract land cover type change patches and corresponding change information. By comparing and analyzing the changes in land cover types and corresponding change information with the farmland project planning scheme, illegal land use after completion can be identified.

3. The farmland construction supervision method according to claim 1, characterized in that, The preset monitoring nodes include the commencement time node, the completion and acceptance node, and at least one phased acceptance node during the construction process.

4. The farmland construction supervision method according to claim 1, characterized in that, Before inputting the high-resolution remote sensing image into a pre-trained land cover classification model for pixel-by-pixel inference, the method further includes: The high-resolution remote sensing image is subjected to standardized preprocessing; the standardized preprocessing includes at least one of radiometric calibration, atmospheric correction and image registration.

5. The farmland construction supervision method according to claim 1, characterized in that, The land feature classification mask is vectorized to obtain land feature attribute vector data, including: In the land feature classification mask, pixels belonging to the same land feature type and spatially connected are aggregated into independent vector plots; Extract the spatial location and vector boundary of each vector vector patch, calculate the geometric area of ​​each vector vector patch, and assign the corresponding land feature type to obtain the land feature attribute vector data.

6. The farmland construction supervision method according to claim 1, characterized in that, The change information of the land feature type change plots includes the land feature type before and after the change, as well as the change in area.

7. The farmland construction supervision method according to claim 6, characterized in that, The changed map patches and their information are compared and analyzed with farmland construction standards to identify illegal construction activities during the construction process, including: Based on the land cover types before and after the change in the map patch, determine whether the change complies with farmland construction standards; If it does not comply with regulations and lacks approval procedures, it will be determined as an illegal construction activity, and the vector boundary of the changed feature will be extracted as the illegal feature.

8. The method for supervising farmland construction according to any one of claims 1-7, characterized in that, The high-resolution remote sensing image is a sub-meter level high-resolution optical remote sensing image.

9. A farmland construction monitoring system, characterized in that, The system is used to implement the farmland construction supervision method as described in any one of claims 1-8, the system comprising: The image acquisition module is used to acquire high-resolution remote sensing images of the farmland project area at key time nodes, including each preset monitoring node during the farmland construction period. The land cover classification module inputs the high-resolution remote sensing image into a pre-trained land cover classification model for pixel-by-pixel inference to obtain the land cover classification mask of the farmland project area at the corresponding key time nodes. The vectorization module is used to vectorize the land feature classification mask to obtain land feature attribute vector data; The feature extraction module is used to perform spatial overlay analysis on the vector data of land cover attributes at different key time points to extract feature type change features and corresponding change information. The progress monitoring module is used to compare and analyze the vector data of the land feature attributes with the vector data of the construction plan to obtain the progress of farmland construction; The violation monitoring module is used to compare and analyze the changed map patches and change information with farmland construction standards to identify illegal construction activities during the construction process.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the farmland construction supervision method according to any one of claims 1-8.

Images