Unmanned aerial vehicle real scene three-dimensional surveying and mapping and plot boundary intelligent replotting method and system for land consolidation

Abstract

The application discloses a land consolidation oriented unmanned aerial vehicle real scene three-dimensional surveying and mapping and land plot boundary intelligent replotting method and system, and belongs to the technical field of land consolidation surveying and mapping and land plot boundary intelligent identification. The method is aimed at the problem that the temporary shovel mark edge and shallow ditch edge in the middle stage of construction are easily misidentified as stable land plot boundaries, generates linear elevation texture candidate boundaries based on unmanned aerial vehicle images and real scene three-dimensional point clouds, constructs cross section disturbance codes, and reconfigures closed land plot boundaries after removing unanchored construction marks in combination with planning anchoring constraints. In the unmanned aerial vehicle real scene three-dimensional surveying and mapping data in the middle stage of land consolidation construction, the method avoids misidentification of the phased construction marks as stable boundary lines that can be used for land plot boundary replotting in the boundary candidate feature extraction link according to the similar trigger conditions of the temporary construction line and the real land plot boundaries in geometry and texture, and improves the stability, traceability and planning consistency of the boundary replotting.

Description

Technical Field

[0001] This invention relates to the field of land consolidation surveying and mapping technology, and more particularly to the field of image recognition. Specifically, it relates to a method and system for UAV real-scene 3D surveying and intelligent redrawing of land boundaries for land consolidation. Background Technology

[0002] Land consolidation is an important engineering approach to optimize the spatial layout of arable land, improve its quality, enhance farmland infrastructure, and promote large-scale agricultural operations. During land consolidation, the accurate identification and rational redrawing of plot boundaries directly affect key aspects such as optimizing plot morphology, maintaining ownership boundaries, laying out irrigation and drainage systems, connecting roads and ditches, and subsequent acceptance and management.

[0003] With the development of UAV oblique photogrammetry, orthophoto production, and real-scene 3D point cloud reconstruction technologies, using UAVs to quickly acquire high-resolution images and 3D geometric information of construction areas has become an important technical approach for dynamic monitoring of land consolidation and extraction of land parcel boundaries. Compared with traditional manual measurement and 2D map interpretation, UAV real-scene 3D mapping can simultaneously express abrupt changes in surface elevation, micro-topographical variations, texture differences, and the spatial morphology of linear features, providing a richer data foundation for intelligent identification of land parcel boundaries and refined management of the construction process.

[0004] However, during the mid-stage of land consolidation construction, numerous continuous shovel marks formed by bulldozing operations and shallow trench edges temporarily excavated for drainage or construction organization will appear at the construction site. These temporary construction lines may exhibit continuous elevation abrupt changes in point clouds and stable grayscale texture differences in orthophotos, and their orientation may be close to the planned field boundary direction. Existing methods for extracting candidate boundaries for land parcels typically use continuous linear elevation abrupt changes, texture segmentation edges, or linear feature outlines as candidate boundary lines. The implicit premise is that the continuous linear landforms visible during surveying already correspond to stable field ridges, ditches, road edges, or ownership boundaries.

[0005] The edges of temporary shovel marks and shallow trenches during the actual construction phase only reflect the stage of earthwork operations and do not correspond to the final plot boundaries. When the aforementioned temporary construction lines simultaneously meet the conditions of geometric continuity and image texture edges, existing methods are prone to mismatching them as candidate plot boundary lines. This causes the boundary redrawing results to locally adhere to the construction marks, resulting in the incorrect segmentation of the same plot, local offset of adjacent plot boundaries, and difficulty in reproducing boundary results generated from surveys at different times. Therefore, a UAV-based real-scene 3D surveying and intelligent plot boundary redrawing method is needed that can distinguish between staged construction marks and stable boundary lines during the mid-term of land consolidation construction. Summary of the Invention

[0006] The technical problem this invention aims to solve is: in the mid-stage of land consolidation construction, in the UAV real-scene 3D mapping data, for the triggering condition that the temporary construction line and the real plot boundary are similar in geometry and texture, to avoid the boundary candidate feature extraction process misidentifying the stage construction traces as stable boundary lines that can be used for plot boundary redrawing.

[0007] To address the aforementioned technical problems, this invention provides a method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation, comprising: S1, acquire the pixel values ​​of UAV oblique photogrammetry images, the pixel values ​​of UAV orthophoto images, the 3D coordinate values ​​of real-world 3D point cloud, the vector coordinates of planned field boundaries, and the pre-configured thresholds, and generate a set of candidate boundaries for linear elevation textures that have completed spatial registration; S2, construct the candidate boundary cross-section elevation sequence and candidate boundary cross-section gray sequence along the candidate boundary set of the linear elevation texture that has been spatially registered, and generate the cross-section perturbation encoded candidate boundary set; S3, based on the planned field boundary vector coordinates, the permanent irrigation and drainage ditch design vector coordinates and the field road design vector coordinates, the cross-sectional disturbance coding candidate boundary set is marked with planning anchoring constraints to obtain the candidate boundary set with completed planning anchoring constraint marking; S4. Based on the cross-sectional disturbance coding attributes and candidate boundary planning anchor marks in the candidate boundary set that have completed the planning anchor constraint marking, the boundary with stage construction traces is eliminated, and a set of stable plot boundary redrawing lines that excludes stage construction traces is reconstructed and generated.

[0008] Further, in step S1, based on the pixel values ​​of the UAV oblique photographic image, the pixel values ​​of the UAV orthophoto image, the 3D coordinate values ​​of the real-world 3D point cloud, the vector coordinates of the planned field boundary, and a pre-configured threshold, a set of candidate boundaries for linear elevation textures that have completed spatial registration is generated, including: S11, the pixel values ​​of the UAV oblique photography image, the pixel values ​​of the UAV orthophoto image, and the three-dimensional coordinate values ​​of the real scene three-dimensional point cloud are uniformly projected to the coordinate system where the boundary vector coordinates of the planned field are located, forming a spatially consistent orthophoto point cloud grid. S12, extract point cloud elevation step line segments and orthophoto texture edge line segments from spatially consistent orthophoto point cloud raster to form elevation texture line segment pairs; S13. Match elevation texture line segment pairs according to the principle of minimum distance between line segment endpoints, and remove elevation texture line segment pairs whose length is less than the pre-configured minimum line segment length threshold to generate a set of candidate boundaries for linear elevation textures that have completed spatial registration.

[0009] Further, in step S11, the pixel values ​​of the UAV oblique photographic image, the pixel values ​​of the UAV orthophoto image, and the three-dimensional coordinate values ​​of the real-world 3D point cloud are uniformly projected onto the coordinate system where the boundary vector coordinates of the planned field are located, forming a spatially consistent orthophoto point cloud raster, including: S111, based on the camera intrinsic parameter matrix, performs the back projection conversion of the pixel values ​​of the UAV oblique photographic image and the pixel values ​​of the UAV orthophoto image from the pixel coordinate system to the camera coordinate system. S112, based on the camera extrinsic parameter matrix, completes the rigid body transformation from the camera coordinate system to the surveying engineering coordinate system; S113, Complete the coordinate transformation from the surveying engineering coordinate system to the coordinate system where the boundary vector coordinates of the planned field are located, according to the coordinate system where the boundary vector coordinates of the planned field are located; S114. Establish a spatial index grid covering the vector coordinate range of the boundary of the planned field plots according to a unified raster resolution, and form a spatially consistent orthophoto point cloud raster containing raster elevation values, raster gray values ​​and spatial correspondence of image point clouds.

[0010] Further, in step S12, point cloud elevation step segments and orthophoto texture edge segments are extracted from the spatially consistent orthophoto point cloud raster to form elevation texture segment pairs, including: S121, calculate the horizontal elevation difference and vertical elevation difference according to the numerical difference of the raster elevation values ​​of the spatial index grid cells, and connect the edges of the spatial index grid cells that reach the pre-configured elevation change threshold to form point cloud elevation step segments. S122, calculate the gray-level gradient magnitude and gray-level gradient direction according to the gray-level difference between adjacent spatial index grid cells, and connect the edges of spatial index grid cells that reach the pre-configured gray-level gradient threshold to form orthophoto texture edge segments; S123, establish candidate pairing relationships for line segments based on the spatial proximity between point cloud elevation step line segments and orthophoto texture edge line segments; S124, write the point cloud elevation step line segments that are spatially adjacent and whose main direction angles meet the pre-configured direction consistency conditions and the orthophoto texture edge line segments into the same elevation texture line segment pair.

[0011] Further, in step S2, constructing a candidate boundary cross-section elevation sequence and a candidate boundary cross-section grayscale sequence along the candidate boundary set of the spatially registered linear elevation texture includes: S21. Along the tangent direction of each candidate boundary in the set of candidate boundaries of the linear elevation texture that has been spatially registered, generate a sequence of point cloud cross-sections perpendicular to the candidate boundaries according to a fixed sampling interval, forming a sequence of candidate boundary cross-section elevations. S22, calculate the center height difference, the slope sign on both sides, the bottom width of the trench and the mean difference of gray values ​​on both sides based on the candidate boundary cross section high program sequence and the candidate boundary cross section gray value sequence, and determine the cross section disturbance coding attributes; S23, write the cross-sectional perturbation coding attribute into the corresponding candidate boundary attribute in the linear elevation texture candidate boundary set that has completed spatial registration, and generate the cross-sectional perturbation coding candidate boundary set.

[0012] Further, in step S22, the center height difference, the slope sign on both sides, the bottom width of the trench, and the mean difference of gray values ​​on both sides are calculated based on the candidate boundary cross-section height sequence and the candidate boundary cross-section gray-scale sequence, and the cross-section perturbation coding attributes are determined, including: S221, calculate the scraper edge determination result according to the single-sided continuous slope break rule, and encode the candidate boundary that satisfies the scraper edge determination result as the scraper edge perturbation code; S222, calculate the shallow ditch edge determination result according to the rule of double-sided slope folding and holding the bottom of the depression, and encode the candidate boundary that satisfies the shallow ditch edge determination result as the shallow ditch edge perturbation code; S223, determine the non-construction disturbance determination result based on the shovel mark edge determination result and the shallow trench edge determination result, and encode the candidate boundary that satisfies the non-construction disturbance determination result as the non-construction disturbance code.

[0013] Further, in step S3, the cross-sectional disturbance coding candidate boundary set is marked with planning anchorage constraints to obtain a candidate boundary set with completed planning anchorage constraint marking, including: S31, each candidate boundary in the cross-sectional disturbance coding candidate boundary set is matched with the nearest distance vector coordinates of the planned field boundary, the design vector coordinates of the permanent irrigation and drainage ditch, and the design vector coordinates of the field road. The endpoint anchorage distance from the endpoint of the candidate boundary to the planned node and the ratio of the planned fitting length between the candidate boundary and the planned line segment are calculated. S32, Generate candidate boundary planning anchor marks according to the rule that the endpoint anchorage distance is not greater than the pre-configured endpoint anchorage distance threshold and the planned fitting length ratio is not less than the pre-configured planned fitting ratio threshold. S33, according to the candidate boundary planning anchorage mark and the connectivity relationship of adjacent candidate boundary endpoints, update the cross-section disturbance coding candidate boundary set to the candidate boundary set that has completed the planning anchorage constraint mark.

[0014] Further, in step S4, based on the cross-sectional disturbance coding attributes and the planned anchorage marks of the candidate boundaries in the completed planning anchorage constraint mark set, the boundary lines with staged construction traces are eliminated, and a set of stable plot boundary redrawing lines excluding staged construction traces is reconstructed and generated, including: S41, Based on the candidate boundary set with cross-sectional disturbance coding attribute of scraper edge disturbance code and candidate boundary planning anchorage mark as unanchored in the candidate boundary set with completed planning anchorage constraint mark, determine the scraper edge removal line segment set; S42, Based on the candidate boundary set with cross-sectional disturbance coding attribute of shallow ditch edge disturbance code and candidate boundary planning anchorage mark of unanchored state in the candidate boundary set with completed planning anchorage constraint mark, determine the set of shallow ditch edge elimination line segments; S43, delete the set of scraper edge removal line segments and the set of shallow ditch edge removal line segments from the candidate boundary set that has completed the planning anchoring constraint marking, and reconstruct the closed plot boundary according to the connection relationship of the endpoints of the candidate boundary, and generate a set of stable plot boundary redrawing lines that excludes stage construction traces.

[0015] Further, in step S43, the set of scraper edge removal segments and the set of shallow trench edge removal segments are deleted from the candidate boundary set after the planning anchoring constraint marking is completed, and the closed plot boundary is reconstructed according to the connection relationship of the endpoints of the retained candidate boundaries, generating a set of stable plot boundary redrawing lines that excludes stage construction traces, including: S431, merge the candidate boundary unique identifiers in the set of scraper edge removal segments and the candidate boundary unique identifiers in the set of shallow trench edge removal segments to form a construction trace removal identifier index; S432, delete the candidate boundaries that hit the construction trace removal identifier index from the candidate boundary set that has completed the planning anchoring constraint mark, and construct the remaining candidate boundary connectivity graph based on the candidate boundary node coordinates and the connectivity relationship of adjacent candidate boundary endpoints of the candidate boundary that has not been deleted; S433, merge endpoints with the same name according to the coincidence of endpoint coordinates, and trace closed loops according to the line segment topology; S434. Based on the self-intersection check results and the area zero value check results, determine the effective closed loops and write the effective closed loops into the set of stable plot boundary redrawing lines that excludes stage construction traces.

[0016] This invention also provides a system for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation, comprising: a data registration and candidate boundary generation module, a cross-section disturbance coding module, a planning anchoring constraint marking module, a construction trace removal and boundary reconstruction module, a storage unit, and a controller; the storage unit stores pixel values ​​of UAV oblique photogrammetry images, pixel values ​​of UAV orthophoto images, 3D coordinate values ​​of real-scene 3D point cloud, vector coordinates of planned field boundaries, design vector coordinates of permanent irrigation and drainage ditches, design vector coordinates of field roads, pre-configured thresholds, a set of candidate boundaries for linear elevation textures that have completed spatial registration, a set of candidate boundaries for cross-section disturbance coding, a set of candidate boundaries that have completed planning anchoring constraint marking, and a set of stable land parcel boundary redrawing lines that exclude staged construction traces; the controller is connected to the data registration and candidate boundary generation module, the cross-section disturbance coding module, the planning anchoring constraint marking module, the construction trace removal and boundary reconstruction module, and the storage unit, and is configured to execute the above-mentioned UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation.

[0017] The beneficial effects of this invention are as follows: By unifying UAV imagery, real-world 3D point clouds, and planning and design vector coordinates into the same spatial index grid, elevation abrupt changes and texture edge features can be used together at the same geographical location, reducing candidate boundary deviations caused by inconsistencies in multi-source data coordinates; by sampling candidate boundary cross-sections and encoding disturbances, temporary shovel marks, temporary ditch edges, and non-construction disturbances can be distinguished, enabling the identification of phased construction traces that are geometrically and texturally similar to real boundaries; by using planning anchoring constraint markers, the spatial fit between candidate boundaries and planned field boundaries, permanent irrigation and drainage ditches, and field roads can be included in the judgment, avoiding the continued use of temporary construction lines lacking planning support for boundary redrawing; by eliminating unanchored shovel marks and ditch edges and reconstructing closed plot boundaries, boundary mis-adsorption and mis-segmentation can be reduced, improving the continuity, stability, and traceability of plot boundary redrawing results. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of candidate boundary extraction for linear elevation texture during the mid-term of land consolidation construction.

[0019] Figure 2 This is a schematic diagram of candidate boundary cross-section sampling and perturbation coding.

[0020] Figure 3 This is a schematic diagram showing the boundary redrawing effect before and after the removal of construction traces.

[0021] Figure 4 It is a flowchart of the technical solution processing procedure. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. These specific embodiments are only for explaining the present invention and do not limit the scope of protection of the present invention.

[0023] Example 1:

[0024] This embodiment provides a method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation. The steps are as follows: Figure 4 As shown, it includes: S1, acquire the pixel values ​​of UAV oblique photogrammetry images, the pixel values ​​of UAV orthophoto images, the 3D coordinate values ​​of real-world 3D point cloud, the vector coordinates of planned field boundaries, and the pre-configured thresholds, and generate a set of candidate boundaries for linear elevation textures that have completed spatial registration; S2, construct the candidate boundary cross-section elevation sequence and candidate boundary cross-section gray sequence along the candidate boundary set of the linear elevation texture that has been spatially registered, and generate the cross-section perturbation encoded candidate boundary set; S3, based on the planned field boundary vector coordinates, the permanent irrigation and drainage ditch design vector coordinates and the field road design vector coordinates, the cross-sectional disturbance coding candidate boundary set is marked with planning anchoring constraints to obtain the candidate boundary set with completed planning anchoring constraint marking; S4. Based on the cross-sectional disturbance coding attributes and candidate boundary planning anchor marks in the candidate boundary set that have completed the planning anchor constraint marking, the boundary with stage construction traces is eliminated, and a set of stable plot boundary redrawing lines that excludes stage construction traces is reconstructed and generated.

[0025] like Figure 1 As shown, during the mid-stage of land consolidation construction, planned field boundaries, permanent irrigation and drainage ditches, field roads, temporary shovel marks, and temporary drainage ditch edges may simultaneously exist within the construction area. UAV orthophotos are used to provide grayscale texture variations, real-world 3D point clouds are used to provide abrupt changes in surface elevation, and the vector coordinates of planned field boundaries are used to provide design spatial constraints for the target boundaries. Figure 1 The solid orange line represents the elevation step line identified by the abrupt change in point cloud elevation; the dashed blue line represents the texture edge line identified by the grayscale texture change in orthophoto; the thick green line represents the candidate boundary formed after the two types of line segments are matched; and the semi-transparent red area represents the temporary shovel marks and temporary shallow trenches formed during the construction period.

[0026] During implementation, the system receives pixel values ​​from UAV oblique photogrammetry images, UAV orthophoto images, 3D coordinates of real-world 3D point clouds, vector coordinates of planned field boundaries, vector coordinates of permanent irrigation and drainage ditches, vector coordinates of field roads, camera intrinsic matrix, camera extrinsic matrix, and pre-configured thresholds. Pixel values ​​from UAV oblique photogrammetry images include pixel row number, pixel column number, and pixel grayscale value; pixel values ​​from UAV orthophoto images include orthophoto pixel row number, orthophoto pixel column number, and orthophoto pixel grayscale value; 3D coordinates of real-world 3D point clouds include the abscissa, ordinate, and elevation coordinates of the point cloud plane; and vector coordinates of planned field boundaries include the coordinates of planned boundary nodes, the sequence of planned boundary nodes, and the topological relationship of planned boundary line segments.

[0027] In step S1, spatial registration and candidate boundary generation are performed on the acquired coordinates: the row and column numbers of the pixel values ​​in the UAV oblique photographic image are used as homogeneous pixel coordinates, the column number is used as the horizontal pixel coordinate, and the row number is used as the vertical pixel coordinate. A back-projection transformation from the pixel coordinate system to the camera coordinate system is then performed based on the camera intrinsic parameter matrix. ; in, This represents the three-dimensional coordinate vector of the pixel values ​​in the camera coordinate system of a drone's oblique photographic image. This represents the scale factor determined by the camera imaging depth or the corresponding point cloud depth. Represents the camera intrinsic parameter matrix. This represents a homogeneous pixel coordinate vector consisting of the pixel column number, pixel row number, and the constant 1.

[0028] For the three-dimensional coordinate vector in the camera coordinate system after back projection transformation Perform a rigid body transformation from the camera coordinate system to the surveying engineering coordinate system: ; in, This represents the pixel coordinates of the UAV oblique photogrammetry image in the surveying engineering coordinate system. This represents the rotation matrix in the camera extrinsic matrix. This represents the translation vector in the camera extrinsic matrix.

[0029] Simultaneously, the orthophoto row number and orthophoto column number of the UAV orthophoto image pixel value are used as homogeneous pixel coordinates, and the homogeneous pixel coordinates are transformed to the surveying and mapping engineering coordinate system to obtain the UAV orthophoto image pixel coordinates in the surveying and mapping engineering coordinate system; the coordinate system in which the real scene 3D point cloud 3D coordinate value is located is the surveying and mapping engineering coordinate system. The 3D coordinates of the real-world 3D point cloud, the pixel coordinates of the UAV oblique photogrammetric image after rigid body transformation, and the pixel coordinates of the UAV orthophoto image are all transformed to the coordinate system of the planned field boundary vector coordinates. Specifically, when a planar affine transformation is used between the surveying engineering coordinate system and the coordinate system of the planned field boundary vector coordinates, the horizontal and vertical coordinates are uniformly converted according to the transformation parameters; when there is an elevation datum difference, the point cloud elevation coordinates are synchronously corrected according to the configured elevation datum correction.

[0030] Furthermore, the coordinates in the coordinate system containing the boundary vector coordinates of the planned field plots are all planar coordinates, which include lateral and longitudinal coordinates. The grid coverage area is determined based on the minimum and maximum lateral, longitudinal, and vertical coordinates of the boundary vector coordinates of the planned field plots, and a uniform raster resolution is extended outside the grid coverage area as a boundary buffer range. As an embodiment of the present invention, the uniform raster resolution is determined based on the ground resolution of the UAV orthophoto and the average point spacing of the real-world 3D point cloud: when the ground resolution of the orthophoto is 0.05m and the average point spacing of the point cloud is not greater than 0.10m, the uniform raster resolution is 0.10m; when the ground resolution of the orthophoto is between 0.10m and 0.20m and the average point spacing of the point cloud is not greater than 0.30m, the uniform raster resolution is 0.20m.

[0031] Based on the unified raster resolution, the coordinate system of the boundary vector coordinates of the planned fields is gridded, and the grid row and column indices of the spatial index grid cells where the coordinates are located after coordinate transformation are generated. The formula for calculating the grid row and column indices is as follows: ; in, The column index representing the spatial index grid cell. This represents the row index of a spatial index grid cell. This represents the horizontal coordinate after coordinate transformation. This represents the vertical coordinate after coordinate transformation. The smallest horizontal coordinate representing the grid coverage area. The smallest vertical coordinate representing the grid coverage area. This indicates a uniform raster resolution.

[0032] For each spatial index grid cell, the 3D coordinates of the real-world 3D point cloud, the pixel values ​​of the UAV oblique photogrammetry image, and the pixel values ​​of the UAV orthophoto image falling into that spatial index grid cell are counted, and an orthophoto point cloud raster is generated. The orthophoto point cloud raster includes the grid row and column index, raster elevation value, and raster grayscale value of the spatial index grid cell.

[0033] Specifically, when the number of 3D coordinate values ​​of the real scene 3D point cloud in the same spatial index grid cell is greater than three, the elevation value of the 3D coordinate values ​​of the real scene 3D point cloud is extracted, and the median elevation value after removing outliers is taken as the raster elevation value of the spatial index grid cell; when the number of 3D coordinate values ​​of the real scene 3D point cloud in the same spatial index grid cell is one to three, the arithmetic mean of the elevation values ​​is taken as the raster elevation value of the spatial index grid cell; when there are no 3D coordinate values ​​of the real scene 3D point cloud in the same spatial index grid cell, the spatial index grid cell is marked as an empty elevation cell and the raster elevation value is set to 0.

[0034] Furthermore, for each spatial index grid cell, the gray values ​​of orthorectified pixels falling into that spatial index grid cell are counted and a raster gray value is generated. When multiple orthorectified pixel gray values ​​exist within the same spatial index grid cell, the effective gray values ​​are selected from the nearest to the farthest according to the distance from the pixel center to the grid cell center (by default, orthorectified pixel gray values ​​with a distance of less than 10 pixels are selected as effective gray values), and the median of the effective gray values ​​is used as the raster gray value.

[0035] Subsequently, elevation texture line segment pairs are extracted from the spatially consistent orthophoto point cloud raster: the orthophoto point cloud raster is traversed according to the row and column order of the spatial index grid cells, and the elevation difference is calculated only for valid locations where there are no empty elevation cell markers in the current spatial index grid cell and its horizontal and vertical adjacent grid cells. ; in, Represents spatial index grid cells Lateral elevation difference between adjacent grid cells Represents spatial index grid cells The longitudinal elevation difference between the grid cells and their adjacent cells. Represents spatial index grid cells The grid elevation value, This represents the row index of a spatial index grid cell. This represents the column index of the spatial index grid cell.

[0036] When the lateral or longitudinal elevation difference reaches a pre-configured elevation change threshold, the common edge of the corresponding grid cell is marked as an elevation step edge point, and the coordinates of the midpoint of this common edge are used as the coordinates of the elevation step edge point. By establishing an adjacency index for all elevation step edge points according to the coordinates of the spatial indexed grid cells, elevation step edge points located at adjacent grid edges and with continuous orientation are written into the same point cloud elevation step segment.

[0037] Specifically, the elevation change threshold is set to a length-type value within the range of 0.10m to 0.30m; when the elevation difference of field ridges, ditch edges, or slopes in the land consolidation survey area is significant and the point cloud elevation accuracy is not less than 0.05m, the elevation change threshold is set to 0.12m; when the surface undulation of the survey area is relatively gentle or the point cloud noise is large, the elevation change threshold is set to 0.20m to 0.30m. Traverse the orthophoto point cloud raster according to the row and column order of the spatial index grid cells, and calculate the gray-level change components based on the horizontally adjacent gray-level values ​​and the vertically adjacent gray-level values: ; in, Represents spatial index grid cells The horizontal grayscale variation component at that location. Represents spatial index grid cells The vertical grayscale variation component at that location. Represents spatial index grid cells The raster grayscale value.

[0038] Calculate the gray-level gradient magnitude and direction based on the horizontal and vertical gray-level variation components: ; in, Represents spatial index grid cells grayscale gradient magnitude at that location Represents spatial index grid cells The direction of the gray-level gradient at that location This represents the two-parameter arctangent function.

[0039] When the grayscale gradient magnitude reaches the pre-configured grayscale gradient threshold, the edge of the corresponding spatial index grid cell is marked as a texture edge point, and the edge direction of the texture edge point is determined according to the grayscale gradient direction. A texture edge adjacency index is established for all texture edge points according to spatial adjacency relationship, and texture edge points located at adjacent grid edges, with continuous grayscale gradient directions and an edge direction angle not exceeding 45 degrees are connected to form orthophoto texture edge line segments.

[0040] Specifically, the grayscale gradient threshold is set to a value in the range of 8 to 25 grayscale levels; when the orthophoto is uniformly illuminated and the grayscale contrast is clear, the grayscale gradient threshold is set to 10 grayscale levels; when the orthophoto has shadows or crop cover texture interference, the grayscale gradient threshold is set to 18 to 25 grayscale levels.

[0041] Point cloud elevation step segments and orthophoto texture edge segments are written into a segment index structure within the same spatial index range. The segment object records the segment's unique identifier, node coordinate sequence, start and end grid cell coordinates, direction, source type, and extraction threshold information. The envelope of each point cloud elevation step segment is expanded by one to three uniform raster resolutions as the nearest neighbor search range to retrieve orthophoto texture edge segments. When an orthophoto texture edge segment and a point cloud elevation step segment are spatially adjacent, and the angle between their principal directions is no greater than 30 degrees, the point cloud elevation step segment and the orthophoto texture edge segment are written into the same elevation texture segment pair. Furthermore, elevation texture segment pairs are matched according to the principle of minimum endpoint distance, and pairs with lengths less than a pre-configured minimum segment length threshold are removed, forming a set of candidate linear elevation texture boundaries with complete spatial registration.

[0042] The minimum line segment length threshold is set to 0.50m to 2.00m. Specifically, when the ground resolution of the UAV orthophoto is no greater than 0.10m and the continuity of the point cloud elevation step line segment is good, the minimum line segment length threshold is 0.50m; when there are fragmented texture edges or local noise in the point cloud in the survey area, the minimum line segment length threshold is 1.00m to 2.00m. Further, in step S13, the unique identifier of the point cloud elevation step segment, the node coordinate sequence of the point cloud elevation step segment, the coordinates of the first endpoint of the point cloud elevation step segment, the coordinates of the last endpoint of the point cloud elevation step segment, the unique identifier of the orthophoto texture edge segment, the node coordinate sequence of the orthophoto texture edge segment, the coordinates of the first endpoint of the orthophoto texture edge segment, the coordinates of the last endpoint of the orthophoto texture edge segment, and the grid row and column index of the corresponding spatial index grid cell are read in each elevation texture segment pair.

[0043] For each pair of elevation texture line segments, the validity of the endpoint coordinates is verified. If the node coordinate sequence of the point cloud elevation step line segment is less than two nodes or the node coordinate sequence of the orthophoto texture edge line segment is less than two nodes, the corresponding elevation texture line segment pair is written into the invalid line segment pair record and does not participate in endpoint matching. When both line segments have valid start and end points, the distance between endpoints in the same direction and the distance between intersection endpoints are calculated respectively. ; in, This represents the distance between the point cloud elevation step line segment and the orthophoto texture edge line segment, formed by aligning the first endpoint with the first endpoint and the last endpoint with the last endpoint. This represents the distance between the intersection points formed by the point cloud elevation step line segment and the orthophoto texture edge line segment, where the first endpoint corresponds to the last endpoint and the last endpoint corresponds to the first endpoint. This represents the coordinate vectors of the first and last endpoints of the point cloud elevation step segment. This represents the coordinate vector of the end point of the point cloud elevation step segment. This represents the coordinate vector of the first and last endpoints of the edge line segment of the orthophoto texture. This represents the coordinate vector of the endpoint of the edge line segment of the orthophoto texture. This represents the L2 norm.

[0044] The matching relationship between the point cloud elevation step line segment and the orthophoto texture edge line segment is determined according to the principle of minimizing the distance between the endpoints of the line segment. The matching direction determination value is determined according to the following formula: ; in, This represents the line segment matching direction determination value. When the line segment matching direction determination value is 0, the same-direction matching state is written; when the line segment matching direction determination value is 1, the coordinate sequence of the line segment nodes of the orthophoto texture edge is reversed and written into the reverse matching state.

[0045] Candidate boundary node coordinates are compiled based on the point cloud elevation step segment node coordinate sequence after direction correction and the orthophoto texture edge segment node coordinate sequence. Specifically, when two line segments have different numbers of nodes and inconsistent node arc length positions, linear interpolation is performed on the other line segment according to the arc length position of the denser node sequence, so that the same arc length position has one set of point cloud elevation step segment node coordinates and one set of orthophoto texture edge segment node coordinates; the midpoint of the two sets of node coordinates at the same arc length position is taken as the candidate boundary node coordinates, and all candidate boundary node coordinates are arranged in ascending order of arc length to form the candidate boundary node coordinate sequence. The matching line segment length is calculated according to the following formula: ; in, Indicates the first The length of the matching line segment formed by the pair of elevation texture line segments. Indicates the first The number of candidate boundary nodes formed by pairs of elevation texture line segments. Indicates the first The first group of elevation texture line segments formed by the first group of elevation texture line segments Coordinate vectors of candidate boundary nodes Indicates the first The first group of elevation texture line segments formed by the first group of elevation texture line segments Coordinate vectors of candidate boundary nodes.

[0046] When the length of the matched line segment is less than the minimum line segment length threshold, the corresponding elevation texture line segment pair is removed from the matching relationship, and the removal record is written with the elevation texture line segment pair identifier, the matched line segment length, the minimum line segment length threshold, the distance between endpoints in the same direction, the distance between endpoints at intersection, and the reason for removal. When the length of the matched line segment is not less than the minimum line segment length threshold, the corresponding elevation texture line segment pair is written into the candidate boundary object. For the retained elevation texture line segment pairs, a unique identifier for the candidate boundary is set, the candidate boundary node coordinate sequence is written into the candidate boundary node coordinate field, the connection order between adjacent candidate boundary nodes is written into the line segment topology relationship field, the unique identifier of the point cloud elevation step line segment, the unique identifier of the orthophoto texture edge line segment, the line segment matching direction determination value, the distance between endpoints in the same direction, the distance between endpoints at intersection, and the matched line segment length are written into the elevation texture matching relationship field, and the grid row and column indices of the spatial index grid cells participating in the matching are written into the spatial index grid cell coordinate field.

[0047] In step S2, as Figure 2 As shown, cross-sections are constructed along each candidate boundary of the spatially registered linear elevation texture. Sampling points are generated along the tangent direction of the candidate boundary at a fixed sampling interval. At each sampling point, point cloud cross-sections are deployed on both sides of the candidate boundary normal, forming a candidate boundary cross-section elevation sequence and a candidate boundary cross-section grayscale sequence. The fixed sampling interval can be 0.20m to 1.00m; when the uniform raster resolution is 0.10m, the fixed sampling interval can be 0.50m; the cross-section half-width can be 0.50m to 2.00m, and adjusted according to the field scale of field ridges, shovel marks, and shallow ditches. Figure 2 The upper half of the diagram is based on an orthophoto of farmland taken by a drone. Red dashed lines represent candidate boundaries of linear elevation textures that have completed spatial registration, while short yellow lines indicate cross-sectional sampling locations along the normal direction of these candidate boundaries. The lower half illustrates three typical cross-sectional morphologies: shovel marks are characterized by continuous sloping edges on one side; shallow ditch edges are characterized by sloping edges on both sides clamping a low-lying trough bottom; and non-construction disturbance morphology is characterized by gently undulating terrain without construction disturbance features. By jointly determining the center elevation difference, slope changes on both sides, trough bottom width, and grayscale differences on both sides in the elevation sequence of the candidate boundary cross-sections, the candidate boundaries are encoded as shovel mark edge disturbance codes, shallow ditch edge disturbance codes, or non-construction disturbance codes. This provides data support for subsequently removing temporary construction traces and preserving stable plot boundaries.

[0048] For each candidate boundary cross-section elevation sequence, the center elevation difference, the slope signs on both sides, the trench bottom width, and the difference in the mean grayscale values ​​on both sides are calculated. When the scraper mark edge exhibits a continuous slope break on one side, the corresponding candidate boundary is encoded as a scraper mark edge perturbation code; when the shallow trench edge exhibits a slope break on both sides sandwiching a low-lying trench bottom, the corresponding candidate boundary is encoded as a shallow trench edge perturbation code; when neither the continuous slope break on one side nor the slope break on both sides sandwiching a low-lying trench bottom rule is met, the corresponding candidate boundary is encoded as a non-construction perturbation code. The scraper mark edge perturbation code, shallow trench edge perturbation code, and non-construction perturbation code are written into the corresponding candidate boundary attributes in the linear elevation texture candidate boundary set that has completed spatial registration, generating a cross-section perturbation encoded candidate boundary set.

[0049] Specifically, in step S21, the set of candidate boundaries for linear elevation textures that have completed spatial registration and the spatially consistent orthophoto point cloud raster are received. The unique identifier of each candidate boundary, the coordinates of the candidate boundary nodes, and the line segment topology are read. The grid row and column indices, raster elevation values, raster grayscale values, and uniform raster resolution of the spatial index grid cells in the spatially consistent orthophoto point cloud raster are also read. The candidate boundary node coordinates are arranged into a continuous sequence according to the line segment topology. Each node in the boundary node coordinate sequence is expressed using planar coordinates in the coordinate system of the planned field boundary vector coordinates.

[0050] The fixed sampling interval is set to 0.50m, the normal sampling half-width is 2.00m, the normal sampling step size is 0.10m, and the abnormal elevation removal threshold is 0.30m. When the unified grid resolution is 0.10m, the normal sampling step size is consistent with the unified grid resolution; when the unified grid resolution is 0.20m, the normal sampling step size is 0.20m, to ensure that the normal sampling points can stably fall within the effective retrieval range of the spatial index grid cells. The line segment length between adjacent nodes is calculated according to the boundary node coordinate sequence, and the boundary arc length coordinates are accumulated along the candidate boundary centerline. Within the boundary arc length coordinate range, the cross-section center sampling point is generated according to the fixed sampling interval. The interpolation positioning calculation formula for the cross-section center sampling point is as follows: ; in, Indicates the first The planar coordinates of the central sampling point of each cross section Indicates containing the first The coordinates of the first node of the line segment at the center sampling point of each cross section. Indicates containing the first The coordinates of the end node of the line segment at the center sampling point of each cross section Indicates the first The boundary arc length coordinates corresponding to the center sampling points of each cross section This represents the cumulative arc length corresponding to the first node of the line segment. This represents the cumulative arc length corresponding to the last node of the line segment.

[0051] Calculate the candidate boundary tangent direction vector based on the candidate boundary line segment where the cross-section center sampling point is located, and then rotate the candidate boundary tangent direction vector to obtain the candidate boundary normal direction vector: ; in, Indicates the first Candidate boundary tangent direction vectors at the center sampling points of each cross section. Indicates the first Candidate boundary normal direction vectors at the center sampling points of each cross section This represents the component of the candidate boundary tangent direction vector on the horizontal coordinate axis. This represents the component of the candidate boundary tangent direction vector on the vertical coordinate axis.

[0052] Around the center sampling point of each cross-section, point cloud sampling points are generated on both sides along the positive and negative directions of the candidate boundary normal direction vector, respectively, so that the point cloud sampling points on both sides cover the elevation change area near the candidate boundary. The calculation formula for generating the point cloud sampling points on both sides is: ; in, Indicates the first The sampling point at the center of the cross section is in the positive direction of the normal. The planar coordinates of the point cloud sampling points on both sides Indicates the first The sampling point at the center of the cross section is located in the opposite direction of the normal. The planar coordinates of the point cloud sampling points on both sides Indicates the normal sampling step size. Indicates the normal sampling half-width, Indicates the normal sampling sequence number.

[0053] Based on the planar coordinates of the point cloud sampling points on both sides, spatial index grid cells are retrieved, and the point cloud elevation coordinates in the corresponding spatial index grid cells and adjacent spatial index grid cells are read. When the point cloud sampling points on both sides fall into a single valid grid cell, the raster elevation value of that spatial index grid cell is directly read as the sampling elevation value. When the point cloud sampling points on both sides are located between multiple spatial index grid cells, local elevation interpolation is performed based on the planar distance from the point cloud sampling points on both sides to the center of the adjacent spatial index grid cells to obtain the interpolated elevation value corresponding to the coordinates of the point cloud sampling points on both sides. When the number of adjacent valid grid cells is less than three, the point cloud sampling points on both sides are marked as sampling points to be supplemented, and the elevation change trend of adjacent valid sampling points on the same side of the same cross section is used to supplement them.

[0054] Abnormal elevation removal is performed on the point cloud sampling points on both sides after elevation interpolation is completed. A local elevation sequence is formed for adjacent normal sampling points within the same cross-section, and the median elevation of this local elevation sequence is used as a stable reference value. When the absolute difference between the sampled elevation value and the stable reference value exceeds the abnormal elevation removal threshold, the sampled elevation value is marked as an abnormal elevation value. The formula for determining abnormal elevation removal is as follows: ; in, Indicates the first The first cross-section The sampling elevation values ​​of the point cloud sampling points on both sides Indicates surrounding the first The first cross-section A local high-order sequence constructed from two side point cloud sampling points, This indicates the threshold for removing abnormal elevations. Sampling points marked as abnormal elevation values ​​are removed from the valid elevation record, and elevation interpolation is performed again based on adjacent valid sampling points on the same side of the same cross section. When the normal length corresponding to consecutive abnormal elevation values ​​is greater than 0.50m, the invalid marking of that segment of sampling points is retained, and forced completion is not performed using distant sampling points.

[0055] According to the generation order of the cross-section center sampling points, the coordinates of the cross-section center sampling points, the tangent direction vector of the candidate boundary, the normal direction vector of the candidate boundary, the coordinates of the normal forward point cloud sampling points, the coordinates of the normal reverse point cloud sampling points, the normal forward sampling elevation value, the normal reverse sampling elevation value, the abnormal elevation removal mark, and the coordinates of the spatial index grid cell corresponding to each candidate boundary are organized into a candidate boundary cross-section elevation sequence, and a one-to-one correspondence is established between the candidate boundary cross-section elevation sequence and the unique identifier of the candidate boundary.

[0056] Specifically, in step S22, the candidate boundary cross-section high-order sequence and spatially consistent orthophoto point cloud raster are received. The unique identifier of each candidate boundary, the coordinates of the cross-section center sampling point, the normal forward sampling elevation value, the normal reverse sampling elevation value, the abnormal elevation removal flag, the grid row and column index of the spatial index grid cell, the raster elevation value, and the raster grayscale value are read. Cross-sections of the same candidate boundary are arranged into a continuous cross-section sequence according to the generation order of the cross-section center sampling points. The effective sampling points in the normal forward and normal reverse directions of each cross-section are organized into high-order sequences of their respective neighboring areas. When the number of effective sampling points in a single cross-section is less than 70% of the total number of sampling points on that side, the cross-section is marked as a low-confidence cross-section, and only its spatial index grid cell coordinates are retained without participating in the disturbance type determination.

[0057] The center neighborhood width is set to 0.20m, the near-side slope break judgment width is 0.60m, the bottom search half-width is 1.20m, the center height difference disturbance threshold is 0.12m, the minimum continuous slope break length is 0.50m, the minimum bottom width is 0.30m, the maximum bottom width is 1.50m, and the gray-scale mean difference discrimination threshold is 12 gray levels. When the orthophoto image gray-scale value is expressed using a single-channel gray-scale value from 0 to 255, the gray-scale mean difference is directly compared according to the gray-scale value. When the orthophoto image gray-scale value is expressed using a multi-channel gray-scale value, the multi-channel pixel value of the corresponding grid unit is first converted into a single-channel gray-scale value before participating in the calculation.

[0058] For each valid cross section, the elevation value at the center of the cross section is extracted as the center elevation value. The average elevation values ​​of the adjacent neighborhoods in the positive direction and the negative direction of the normal are extracted as the elevation values ​​of the adjacent neighborhoods on both sides. The elevation difference between the center of the cross section is then calculated. ; in, Indicates the first The center elevation difference of each cross section Indicates the first The center elevation value of each cross section, Indicates the first The average elevation of the nearest neighbor region in the positive direction of the cross-sectional normal. Indicates the first The average elevation of the adjacent area in the opposite direction of the cross-sectional normal.

[0059] The elevation increase / decrease relationship of adjacent sampling points is calculated from the center outwards according to the normal-direction sampling points, forming a normal-direction slope symbol sequence; the elevation increase / decrease relationship of adjacent sampling points is calculated from the center outwards according to the normal-reverse sampling points, forming a normal-reverse slope symbol sequence. Sampling intervals with an elevation difference greater than 0.03m between adjacent sampling points are marked as rising symbols, those with an elevation difference less than -0.03m are marked as falling symbols, and those with an absolute elevation difference of no more than 0.03m between adjacent sampling points are marked as level symbols. The lengths of consecutive slope symbols on the same side within continuous cross-sections of the same candidate boundary are also recorded. When consecutive rising or falling symbols appear on either the normal-direction or normal-reverse side, and the duration of this continuous slope break in the candidate boundary direction is not less than 0.50m, and the absolute value of the elevation difference between the cross-section centers is not less than 0.12m, the corresponding cross-section segment of the candidate boundary is marked as a single-sided continuous slope break segment.

[0060] For each valid cross-section, a set of sampling points continuously lower than the elevations of the adjacent adjacent areas is searched within the half-width range of the trench bottom. The continuous coverage length of this set of sampling points in the normal direction is taken as the trench bottom width, and the elevation values ​​of this continuous set of sampling points must be at least 0.08m lower than the smaller of the average elevation values ​​of the adjacent adjacent areas. The formula for calculating the trench bottom width is: ; in, Indicates the first The width of the bottom of the trench in each cross section, Indicates the first The number of consecutive low-lying sampling points in a cross-section This indicates the normal sampling step size. When the signs of the positive and negative slopes of the normal change in opposite directions in the same cross section, and the width of the trench bottom is not less than 0.30m and not more than 1.50m, the cross section is marked as a cross section with a low-lying trench bottom sandwiched by two slopes. When cross sections with a low-lying trench bottom sandwiched by two slopes appear continuously in the same candidate boundary direction, and the continuous length is not less than 0.50m, the corresponding candidate boundary section is marked as a shallow ditch morphology section.

[0061] Based on the coordinates of the central sampling point of each cross section and the sampling points on both sides of its normal, grayscale raster cells in the spatially consistent orthophoto point cloud are retrieved. The grayscale sampling values ​​of the normal's forward neighborhood and the normal's reverse neighborhood are read respectively. Invalid sampling points corresponding to abnormal elevation removal marks are removed, and the difference in grayscale mean between the two sides is calculated. ; in, Indicates the first The difference in mean grayscale values ​​between the two sides of a cross-section. Indicates the first The average gray value of the positive neighborhood of the cross-sectional normal. Indicates the first The average gray value of the reverse neighborhood of the cross-sectional normal. This indicates taking the absolute value.

[0062] The center elevation difference, slope signs on both sides, trench bottom width, and mean gray value difference on both sides of all valid cross sections within the same candidate boundary are merged and judged according to the candidate boundary direction. When a candidate boundary has a continuous slope break section on one side, and the mean gray value difference on both sides of the section reaches more than 12 gray levels, without forming a low-lying trench bottom shape sandwiched by double slope breaks, the disturbance type of the candidate boundary is written into the shovel mark edge disturbance code, and the corresponding center elevation difference range, continuous slope break direction, continuous slope break length, and mean gray value difference range are recorded. When a candidate boundary has a shallow ditch section, and the trench bottom width in the section is between 0.30m and 1.50m, and the slope signs on both sides change in opposite directions and remain stable in the continuous cross section, the disturbance type of the candidate boundary is written into the shallow ditch edge disturbance code, and the corresponding trench bottom width range, low-lying depth range, double slope break direction, and continuous shallow ditch length are recorded. When a candidate boundary does not form a continuous single-sided slope bend section, nor does it form a cross-section of a low-lying trench bottom sandwiched by double-sided slope bends, or although there are local elevation fluctuations, the continuous length is less than 0.50m, the central elevation difference is less than 0.12m, and the trench bottom width is not within the range of 0.30m to 1.50m, the disturbance type of the candidate boundary is written into the non-construction disturbance code, and its central elevation difference, slope sign, trench bottom width, and grayscale mean difference are retained as disturbance feature records. A one-to-one correspondence is established between the disturbance code and the candidate boundary according to the unique identifier of the candidate boundary, and the candidate boundary disturbance coding result, which includes the unique identifier of the candidate boundary, the shovel mark edge disturbance code, the shallow ditch edge disturbance code, the non-construction disturbance code, and the disturbance feature record, is output.

[0063] Specifically, in step S23, the following steps are taken: receiving the shovel mark edge disturbance code, the shallow ditch edge disturbance code, the non-construction disturbance code, and the set of candidate boundaries for linear elevation textures that have been spatially registered; reading the unique identifier of the candidate boundary, the coordinates of the candidate boundary node, the line segment topology, the elevation texture matching relationship, and the cross-sectional elevation sequence index corresponding to each candidate boundary in the set of candidate boundaries for linear elevation textures that have been spatially registered; and reading the cross-sectional disturbance code value and disturbance feature record corresponding to the unique identifier of the candidate boundary in the candidate boundary disturbance coding result.

[0064] The unique identifier of the candidate boundary is used as the primary key of the attribute mapping. The disturbance codes of the scraper edge, the shallow ditch edge, and the non-construction disturbance codes are organized into a unified cross-sectional disturbance coding field. The effective value range of the cross-sectional disturbance coding field is set to the scraper edge disturbance code, the shallow ditch edge disturbance code, and the non-construction disturbance code. The range of center elevation difference, continuous slope break direction, continuous slope break length, ditch bottom width range, depression depth range, double slope break direction, continuous shallow ditch length, and the range of gray mean difference between the two sides are organized into disturbance feature record fields.

[0065] A candidate boundary object index table is established based on the unique identifier of the candidate boundary, so that each unique identifier of the candidate boundary points to a candidate boundary object in the set of candidate boundaries of linear elevation texture that has been spatially registered. Then, a perturbation coding index table is established based on the same unique identifier of the candidate boundary, so that each unique identifier of the candidate boundary points to a cross-sectional perturbation coding value and a set of perturbation feature records. Consistency checks are performed on the candidate boundary object index table and the perturbation coding index table. When the unique identifier of a candidate boundary exists in both index tables, the corresponding cross-section perturbation coding value is written to the cross-section perturbation coding attribute field of the candidate boundary object, and the corresponding perturbation feature record is written to the perturbation feature record attribute field of the candidate boundary object. When the unique identifier of a candidate boundary exists in the candidate boundary object index table but not in the perturbation coding index table, the candidate boundary node coordinates, line segment topology, elevation texture matching relationship, and cross-section elevation sequence index of the candidate boundary object are retained, and the candidate boundary object is marked as a perturbation coding missing record. When there is a duplicate unique identifier of a candidate boundary in the perturbation coding index table, a valid mapping record is retained according to the latest complete perturbation feature record corresponding to the unique identifier of the candidate boundary, and the duplicate record is written to the attribute update verification record.

[0066] For candidate boundary objects that have completed effective mapping, the coordinate order of candidate boundary node coordinates is not changed, the connection relationship between the first and last points and adjacent nodes in the line segment topology is not changed, the matching identifier between the point cloud elevation step line segment and the orthophoto texture edge line segment in the elevation texture matching relationship is not changed, and the correspondence between the cross-section elevation sequence index and the unique identifier of the candidate boundary is not changed. Only the cross-section perturbation coding attribute, perturbation feature record attribute, and attribute update time stamp are added to the candidate boundary attribute structure. After the attributes are written, the candidate boundary objects are reorganized according to the unique identifier of the candidate boundary. The candidate boundary objects with the cross-section perturbation coding attribute written, the retained candidate boundary node coordinates, line segment topology, elevation texture matching relationship, cross-section elevation sequence index, and attribute update verification record are all written into the updated candidate boundary set to form the cross-section perturbation coded candidate boundary set.

[0067] In step S3, each candidate boundary in the cross-sectional disturbance coding candidate boundary set is matched with the nearest distance vector coordinates of the planned field boundary, the design vector coordinates of the permanent irrigation and drainage ditch, and the design vector coordinates of the field road. The endpoint anchorage distance from the endpoint of the candidate boundary to the planned node and the planned fitting length ratio between the candidate boundary and the planned line segment are calculated.

[0068] Furthermore, when the endpoint anchorage distance is not greater than the pre-configured endpoint anchorage distance threshold and the planned fitting length ratio is not less than the pre-configured planned fitting ratio threshold, the corresponding candidate boundary is marked as anchored; otherwise, the corresponding candidate boundary is marked as unanchored. Based on the planned anchorage markings of the candidate boundaries and the connectivity relationships of adjacent candidate boundary endpoints, the cross-sectional disturbance coding candidate boundary set is updated to the candidate boundary set that has completed the planned anchorage constraint markings.

[0069] The endpoint anchorage distance threshold can be taken from 0.20m to 1.00m. Specifically, when the registration error between the planned design vector coordinates and the UAV survey coordinates is no greater than 0.20m, the endpoint anchorage distance threshold can be taken as 0.30m; when there is significant surface disturbance during construction and local measurement errors exist along the planned line, the endpoint anchorage distance threshold can be taken as 0.50m to 1.00m. The planning fit ratio threshold can be taken from 0.60 to 0.90; when the boundary is required to be at the same height as the planned line, the planning fit ratio threshold can be taken as 0.80; when there is local construction offset in the planned ditch or road, the planning fit ratio threshold can be taken as 0.60 to 0.70.

[0070] Specifically, in step S31, the system receives the cross-sectional disturbance coding candidate boundary set, the planned field boundary vector coordinates, the permanent irrigation and drainage ditch design vector coordinates, and the field road design vector coordinates. It reads the unique identifier of each candidate boundary, the candidate boundary node coordinates, the line segment topology, and the cross-sectional disturbance coding attributes corresponding to each candidate boundary in the cross-sectional disturbance coding candidate boundary set. It also reads the polyline node coordinates, polyline node order, and line segment connection relationships in each planning and design vector coordinate. The planned field boundary vector coordinates, permanent irrigation and drainage ditch design vector coordinates, and field road design vector coordinates are uniformly organized into a planning and design vector element set, while maintaining the source type identifiers of the planned field boundaries, permanent irrigation and drainage ditches, and field roads. Adjacent polyline nodes in each planning and design vector element are connected to form a planning line segment object, and each polyline node is organized into a planning node object, forming a planning line segment index set and a planning node index set, respectively. The grid side length of the spatial retrieval grid is taken as 1.00m to 5.00m; when the density of the planned line segments is high, 1.00m is taken, and when the distribution of the planned line segments is sparse, 5.00m is taken.

[0071] Using the first and last endpoints of the candidate boundary as search centers, spatially adjacent planning node objects are retrieved from the planning node index set, and the distances from the endpoints of the candidate boundaries to the nearest planning node are calculated for each. ; in, This represents the distance from the candidate boundary endpoint to the nearest planned node. Represents the first node in the planning node index set. One planning node, Represents the set of planning node indices. Indicates the lateral coordinates of the candidate boundary endpoints. Represents the vertical coordinates of the candidate boundary endpoints. Indicates the first The horizontal coordinates of each planning node, Indicates the first The longitudinal coordinates of each planning node are determined. The anchorage distances at the first and last endpoints of the candidate boundary are calculated using this method, and the larger of these two values ​​is taken as the endpoint anchorage distance of the candidate boundary. ; in, Indicates the anchorage distance at the endpoints of the candidate boundary. Indicates the anchorage distance at the first endpoint of the candidate boundary. This indicates the anchorage distance at the end point of the candidate boundary.

[0072] The lengths of line segments between adjacent candidate boundary nodes are calculated along the coordinate sequence of candidate boundary nodes, and accumulated to form the total length of candidate boundaries. The fitting sampling interval is 0.20m to 1.00m, and the buffer distance for planned line segments is 0.30m to 1.50m. When the planar resolution of the UAV real-world 3D point cloud is no greater than 0.10m, the fitting sampling interval is 0.20m; when the planar resolution of the UAV real-world 3D point cloud is greater than 0.10m but no greater than 0.50m, the fitting sampling interval is 0.50m; when the scale of the planned line segments in the survey area is large and the point cloud density is low, the fitting sampling interval is 1.00m. Fitting sampling points are generated on the coordinate sequence of candidate boundary nodes according to the fitting sampling interval, and each fitting sampling point is used as the retrieval center to retrieve neighboring planned line segment objects located within the extended range of the buffer distance of the planned line segments from the planned line segment index set.

[0073] For each fitting sampling point, project the fitting sampling point onto the adjacent planned line segment object to obtain the perpendicular foot position and projection distance from the fitting sampling point to the planned line segment: ; in, Indicates the first The first fitting sampling point at the... Projection parameters on the planned line segment, Indicates the first A planar coordinate vector that fits the sampling point. Indicates the first The first node plane coordinate vector of the planned line segment, Indicates the first The planar coordinate vector of the last node of the planned line segment, Indicates the first The first fitting sampling point at the... The projection perpendicular plane coordinate vector on the planned line segment, This indicates selecting 0. The maximum value in, This indicates selecting 1. The minimum value in the calculation is used. When the distance from the fitting sampling point to the nearest planned line segment is not greater than the buffer distance of the planned line segment, the fitting sampling point is marked as a fitting sampling point. The fitting sampling length is counted according to the candidate boundary direction, and the proportion of planned fitting length is calculated. ; in, This indicates the proportion of the planned fit length. Indicates the fit sampling length. This represents the total length of the candidate boundary.

[0074] For candidate boundaries that pass the validity verification, jointly determine whether the candidate boundaries satisfy the planning anchorage constraints: ; in, Indicates the first The planning anchorage determination value of the candidate boundary. Indicates the first The endpoint anchorage distance of the candidate boundary This indicates the pre-configured endpoint anchorage distance threshold. Indicates the first The proportion of the planned fitting length of the candidate boundaries, This indicates the pre-configured planning fit ratio threshold. When the planning anchoring judgment value is 1, the candidate boundary planning anchoring mark is written to the anchored state; when the planning anchoring judgment value is 0, the candidate boundary planning anchoring mark is written to the unanchored state, and the unanchored reason field is written as endpoint not anchored, insufficient fit, or endpoint not anchored and insufficient fit.

[0075] Specifically, in step S33, the unique identifier of the candidate boundary is used as the primary key for attribute mapping to construct a candidate boundary object index table, a planning anchoring mark index table, and a planning geometry matching result index table. When the same unique identifier of the candidate boundary exists in all three index tables, the candidate boundary planning anchoring mark, endpoint anchoring distance, and planning fitting length ratio are determined as the planning anchoring constraint attributes to be written. When the candidate boundary object index table contains the unique identifier of the candidate boundary but the planning anchoring mark index table or the planning geometry matching result index table lacks the unique identifier of the candidate boundary, the candidate boundary node coordinates, line segment topology, cross-section disturbance coding attributes, and disturbance feature record attributes are retained, and the candidate boundary planning anchoring mark is written to the unanchored state.

[0076] According to the line segment topology, the coordinates of the candidate boundary nodes of each candidate boundary are arranged into a continuous node sequence. The first node in the continuous node sequence is determined as the first endpoint of the candidate boundary, and the last node in the continuous node sequence is determined as the last endpoint of the candidate boundary. The endpoint connectivity judgment tolerance is 0.05m to 0.30m; 0.05m is used when the plane resolution of the UAV real-view 3D point cloud is not greater than 0.10m, 0.20m is used when the plane resolution of the UAV real-view 3D point cloud is greater than 0.10m but not greater than 0.50m, and 0.30m is used when there is a slight edge-connection error in the vector coordinates of the planned field boundary. The endpoint connectivity judgment value is calculated according to the following formula: ; in, This represents the connectivity determination value for candidate boundary endpoints. and Indicates the first The first candidate boundary The horizontal and vertical coordinates of each endpoint and Indicates the first The first candidate boundary The horizontal and vertical coordinates of each endpoint This indicates the endpoint connectivity judgment tolerance. Further, when the endpoint connectivity judgment value is 1, the endpoints of the two corresponding candidate boundaries are written to the same endpoint connectivity node, and the unique identifier of adjacent candidate boundaries, adjacent endpoint types, endpoint connectivity distance, and endpoint connectivity node identifier are recorded. The candidate boundary planning anchoring mark, endpoint anchoring distance, planning fit length ratio, nearest planning node source type, main matching source type, and adjacent candidate boundary endpoint connectivity relationship are written to the planning anchoring constraint attribute record of the corresponding candidate boundary object, generating a candidate boundary set with completed planning anchoring constraint markings.

[0077] In step S4, based on the candidate boundaries in the completed planning anchoring constraint marking set that have cross-sectional disturbance coding attributes of shovel mark edge disturbance code and whose planned anchoring marking is in an unanchored state, a set of shovel mark edge removal segments is determined; based on the candidate boundaries in the completed planning anchoring constraint marking set that have cross-sectional disturbance coding attributes of shallow ditch edge disturbance code and whose planned anchoring marking is in an unanchored state, a set of shallow ditch edge removal segments is determined. This process ensures that candidate boundaries identified as temporary construction disturbances and lacking planning design element anchoring support are no longer involved in the stable plot boundary reconstruction.

[0078] Specifically, in step S41, the set of candidate boundaries with completed planning and anchoring constraint markings is received. For each candidate boundary, the unique identifier of the candidate boundary, the coordinates of the candidate boundary nodes, the line segment topology, the cross-sectional disturbance coding attribute, the candidate boundary planning and anchoring mark, and the connectivity of the endpoints of adjacent candidate boundaries are read. Valid values ​​for the cross-sectional disturbance coding attribute are limited to shovel edge disturbance code, shallow trench edge disturbance code, and non-construction disturbance code. Valid values ​​for the candidate boundary planning and anchoring mark are limited to anchored and unanchored states. When reading each candidate boundary, the unique identifier of the candidate boundary, the cross-sectional disturbance coding attribute, and the candidate boundary planning and anchoring mark are checked for field integrity. If any field is missing, the candidate boundary is retained in the set of candidate boundaries with completed planning and anchoring constraint markings, not written to the shovel edge removal line segment set, and recorded as a candidate boundary with missing attributes.

[0079] For candidate boundaries that pass the field integrity verification, the coordinates of candidate boundary nodes, the topological relationships of line segments, and the connectivity relationships of adjacent candidate boundary endpoints remain unchanged. Only the cross-sectional disturbance coding attributes and the planned anchoring marks of the candidate boundaries are evaluated for consistency. The shovel mark edge removal judgment value is generated according to the following formula: ; in, Indicates the first The culling edge rejection value for candidate boundaries. This is an indicator function that takes the value 1 when the condition is true and 0 when the condition is false. AND operation, Indicates the first The cross-sectional perturbation encoding attribute value of the candidate boundary. This indicates the value of the disturbance code identifier for the scraper edge. Indicates the first Candidate boundary planning anchorage marker values ​​for candidate boundaries. This indicates the unanchored state identifier value. When the scraper edge removal judgment value is 1, the corresponding candidate boundary is written into the scraper edge removal segment set, and the unique identifier of the candidate boundary, the coordinates of the candidate boundary node, the topological relationship of the segment, the connectivity relationship of the endpoints of adjacent candidate boundaries, the cross-sectional disturbance coding attribute, and the planned anchoring mark of the candidate boundary are retained; when the scraper edge removal judgment value is 0, the candidate boundary is not written into the scraper edge removal segment set.

[0080] Specifically, in step S42, the set of candidate boundaries with completed planning anchoring constraint markings and the set of shovel edge removal segments are received. The unique identifier of each selected candidate boundary in the shovel edge removal segment set is read, and the read unique identifiers of candidate boundaries are organized into a selected removal identifier index. When the shovel edge removal segment set is empty, an empty selected removal identifier index is created. The unique identifier of the candidate boundary, the coordinates of the candidate boundary nodes, the segment topology, the cross-sectional disturbance coding attribute, the candidate boundary planning anchoring mark, and the connectivity of the endpoints of adjacent candidate boundaries are read from each candidate boundary in the set of candidate boundaries with completed planning anchoring constraint markings. Field integrity verification and valid value verification are performed. Candidate boundaries that fail the verification are not written into the shallow ditch edge removal segment set.

[0081] For candidate boundaries that have passed verification, the coordinates of candidate boundary nodes, the topological relationships of line segments, and the connectivity of adjacent candidate boundary endpoints remain unchanged. Only the unique identifier of the candidate boundary, the cross-sectional disturbance coding attribute, and the planned anchoring mark of the candidate boundary are jointly determined for consistency. The shallow ditch edge removal judgment value is generated according to the following formula: ; in, Indicates the first The shallow groove edge rejection criterion value for candidate boundaries. This indicates the perturbation code identifier value at the edge of the shallow trench. Indicates the first A unique identifier for each candidate boundary. This indicates that it does not belong to [the group / organization]. This indicates the selected elimination index. When the shallow ditch edge elimination judgment value is 1, the corresponding candidate boundary is written into the shallow ditch edge elimination segment set, and the unique identifier of the candidate boundary, the coordinates of the candidate boundary nodes, the topological relationship of the segment, the connectivity relationship of the endpoints of adjacent candidate boundaries, the cross-sectional disturbance coding attribute, the candidate boundary planning anchoring mark, and the status of the selected elimination index not being hit are retained; when the shallow ditch edge elimination judgment value is 0, the candidate boundary is not written into the shallow ditch edge elimination segment set. For the case where the unique identifier of the same candidate boundary appears repeatedly, priority is given to retaining the record where the candidate boundary node coordinates, segment topological relationship, cross-sectional disturbance coding attribute, candidate boundary planning anchoring mark, and connectivity relationship of the endpoints of adjacent candidate boundaries are all complete.

[0082] Furthermore, the unique identifiers of candidate boundaries in the set of removed segments for scraper edge elimination and the set of removed segments for shallow trench edge elimination are merged to form a construction trace elimination identifier index. The set of candidate boundaries for which planned anchoring constraint markings have been completed is traversed. Candidate boundaries that match the construction trace elimination identifier index are removed from subsequent connected graph construction objects. The unique identifier of the candidate boundary, cross-sectional disturbance coding attribute, planned anchoring mark of the candidate boundary, coordinates of the original candidate boundary node, original segment topology, connectivity of the original adjacent candidate boundary endpoints, and elimination source are written into the construction trace elimination record.

[0083] Based on the topological relationship of each candidate boundary line segment, the coordinates of the candidate boundary nodes are arranged into a continuous node coordinate sequence, and the first and last endpoints of the candidate boundaries are determined. The endpoint merging distance tolerance can be between 0.05m and 0.30m; when the endpoint coordinates are recorded with centimeter-level precision and the discrete range of the same-name endpoints is no greater than 0.05m, the endpoint merging distance tolerance is 0.05m; when there are edge discretenesses of more than 0.05m but no more than 0.20m in the endpoint coordinates, the endpoint merging distance tolerance is 0.20m; when there are slight edge errors in the same-name endpoints but the endpoint connectivity can be mutually verified, the endpoint merging distance tolerance is 0.30m.

[0084] For spatially adjacent endpoints that lack endpoint connectivity identifiers or have inconsistent endpoint connectivity identifiers, determine whether they belong to the same connected graph node based on the plane distance between the endpoints: ; in, Indicates the first The reserved candidate boundary endpoints and the first Planar distance between each retained candidate boundary endpoint Indicates the first The lateral coordinates of the candidate boundary endpoints are retained. Indicates the first The longitudinal coordinates of the candidate boundary endpoints are retained. Indicates the first The lateral coordinates of the candidate boundary endpoints are retained. Indicates the first The longitudinal coordinates of each candidate boundary endpoint are retained.

[0085] The merged connected graph nodes are used as nodes in the candidate boundary connected graph. Each candidate boundary is used as an edge connecting the connected graph node containing its first endpoint and the connected graph node containing its last endpoint. The edge attributes record the unique identifier of the candidate boundary, the coordinate sequence of the forward node, the coordinate sequence of the reverse node, the identifier of the connected graph node containing the first endpoint, the identifier of the connected graph node containing the last endpoint, and the original line segment topology. For each edge, a forward directed edge and a reverse directed edge are established, so that closed loops can be traced continuously in any connected graph node.

[0086] At each connected graph node, calculate the orientation order according to the tangent direction of each directed graph edge leaving the node, and establish an ordered index for outgoing edges. For a directed graph edge entering the current connected graph node, take its terminal tangent direction as the incoming edge direction, and take the initial tangent direction of each candidate outgoing edge at the current connected graph node as the outgoing edge direction. Calculate the directed rotation angle between the incoming and outgoing edge directions: ; in, Represents the edges of a directed graph The direction of the incoming edge is turned to the direction of the graph edge. The non-negative directed rotation angle formed by the outgoing edge direction. Represents the edges of a directed graph The component of the tangent direction at the end on the horizontal coordinate axis. Represents the edges of a directed graph The component of the tangent direction at the end on the longitudinal coordinate axis. This represents the component of the tangent direction at the beginning of the edge of the directed graph on the horizontal coordinate axis. Represents the edges of a directed graph The component of the tangent direction at the beginning on the longitudinal coordinate axis. This represents the two-parameter arctangent function. This means normalizing the angle to a value that is not less than zero and less than 0. The range.

[0087] Starting with a directed graph edge that has not yet been marked as a loop, this edge is designated as the starting directed graph edge. Candidate outgoing edges are retrieved along the next connected graph node it points to. Reverse directed graph edges that are completely opposite in direction to the current edge and would immediately return to the previous connected graph node are excluded. Among the remaining candidate outgoing edges, the one with the smallest non-negative directed turning angle is selected as the next directed graph edge. The corresponding candidate boundary node coordinate sequences are then concatenated according to the direction of the selected edge. This process of connecting graph node retrieval, candidate outgoing edge filtering, and node coordinate sequence concatenation continues until the starting connected graph node of the starting directed graph edge is reached again and the next directed graph edge is consistent with the starting directed graph edge. At this point, the concatenated node coordinate sequence is identified as a closed candidate loop.

[0088] For each closed candidate loop, adjacent duplicate nodes are removed from the splicing node coordinate sequence, retaining only the first and last closed nodes. The unique identifiers of all candidate boundaries participating in the tracing of this closed candidate loop are written into the source candidate boundary unique identifier set in the tracing order. Duplicate loop elimination, self-intersection checking, and zero-area checking are performed on the closed candidate loops. The signed area is calculated for closed candidate loops that have not undergone self-intersection. ; in, Indicates the first The signed area of ​​each closed candidate loop. Indicates the first The number of nodes in a closed candidate loop that are not counted repeatedly as the endpoint of the loop. Indicates the first The first closed candidate loop The horizontal coordinates of each node, Indicates the first The first closed candidate loop The vertical coordinates of each node.

[0089] The tolerance for determining zero area can be 0.5 to 2 times the square of the tolerance for the distance between endpoints; when the tolerance for the distance between endpoints is 0.10m, the tolerance for determining zero area is 0.01m. 2 When the endpoint merging distance tolerance is 0.20m, the area zero value determination tolerance is 0.04m. 2 Closed candidate loops with an absolute area value not exceeding the tolerance for zero area are identified as closed loops with zero area and deleted. Closed candidate loops with an absolute area value exceeding the tolerance for zero area are retained as geometrically valid closed loops. Using the horizontal and vertical coordinate axes of the projected coordinate system where the planned field boundary vector coordinates are located as direction references, the outer boundary is uniformly set to a counterclockwise direction. When the signed area of ​​a geometrically valid closed loop is less than zero, the sequence of closed plot boundary nodes, the topological relationship of closed line segments, and the unique identifier set of the source candidate boundaries are synchronously reversed.

[0090] For each geometrically valid and oriented closed parcel boundary, a stable parcel boundary redrawing line object is generated. This object is then used to record the unique identifier of the stable parcel boundary redrawing line, the sequence of closed parcel boundary nodes, the topological relationship of the closed line segments, the set of unique identifiers for the source candidate boundaries, the sequence of node identifiers in the connected graph, the endpoint merging distance tolerance, the signed area, the self-intersection check result, the area zero-value check result, and the boundary direction correction result. Additionally, the unique identifiers of deleted candidate boundaries adjacent to or related to the source candidate boundaries and their original connected relationships, the source of removal, the cross-sectional disturbance coding attribute, the candidate boundary planning anchorage mark, and the invalid adjacent references are written into the corresponding construction trace removal record. This results in a set of stable parcel boundary redrawing lines excluding staged construction traces. For example... Figure 3 The figure shows the boundary redrawing effect before and after the removal of construction traces. The left side shows the result before processing. Temporary shovel marks and shallow ditches were incorrectly identified as candidate boundaries due to both abrupt changes in point cloud elevation and differences in image texture, resulting in the missegmentation of local plots. The right side shows the result after processing. The red dashed line represents the construction traces identified and removed through cross-sectional disturbance coding, the thick green solid line represents the retained stable boundaries, and the cyan dashed line represents the connection relationship between the stable boundaries and the planned boundaries, ditches, or roads. This figure demonstrates that this application can effectively distinguish between temporary construction disturbances and stable plot boundaries during the mid-term of land remediation construction, avoiding false boundary adsorption and missegmentation, and improving the continuity, stability, and planning consistency of the plot boundary redrawing results.

[0091] In another embodiment, when there are shadows, color differences in bare soil, or reflections from localized water accumulation in the orthophoto of the UAV, the grayscale gradient threshold can be selected from 18 to 25 grayscale levels, and the judgment weight of the difference in grayscale mean on both sides is increased in the cross-sectional disturbance coding; this replacement does not change the overall inventive concept of eliminating staged construction traces based on elevation texture candidate boundaries, cross-sectional disturbance coding, and planned anchoring constraints.

[0092] In another embodiment, when the permanent irrigation and drainage ditches or field roads in the survey area have been partially formed in the middle of construction but have slight deviations from the planned line, the planning fit ratio threshold can be taken as 0.60 to 0.70, and the candidate boundary can be judged to have planning support in combination with the endpoint anchorage distance threshold. This replacement can retain stable linear boundaries with planning basis, while still eliminating the edges of temporary shovel marks and shallow ditches that are not anchored.

[0093] Example 2: As an embodiment of the present invention, a system for real-scene 3D mapping and intelligent redrawing of land parcel boundaries using UAVs for land consolidation is provided, including a data registration and candidate boundary generation module, a cross-section disturbance coding module, a planning anchoring constraint marking module, a construction trace removal and boundary reconstruction module, a storage unit, and a controller.

[0094] Specifically, the storage unit stores pixel values ​​of UAV oblique photogrammetry images, pixel values ​​of UAV orthophoto images, 3D coordinate values ​​of real-world 3D point clouds, vector coordinates of planned field boundaries, design vector coordinates of permanent irrigation and drainage ditches, design vector coordinates of field roads, pre-configured thresholds, and data sets generated at each processing stage. The controller is connected to each functional module (including a data registration and candidate boundary generation module, a cross-sectional disturbance coding module, a planning anchoring constraint marking module, and a construction trace removal and boundary reconstruction module) and the storage unit, and is used to call various types of surveying and mapping data and control each module to execute the land consolidation area boundary identification and redrawing process. During system operation, the data registration and candidate boundary generation module obtains UAV oblique photogrammetry images, orthophoto images, and real-world 3D point cloud data from the storage unit. Through spatial coordinate transformation and 3D point cloud matching processing, it achieves spatial registration of multi-source surveying and mapping data, and extracts linear spatial objects with boundary features within the land consolidation area based on the registered elevation change characteristics, generating a set of linear elevation texture candidate boundaries that have completed spatial registration, providing a data foundation for subsequent candidate boundary stability analysis.

[0095] The system uses a cross-section perturbation coding module and a planning anchoring constraint marking module to realize the spatial feature analysis of candidate boundaries and the association with planning constraints. Specifically, the cross-section perturbation coding module samples along the set of linear elevation texture candidate boundaries that have completed spatial registration, constructs the corresponding cross-section analysis region according to the spatial orientation of the candidate boundaries, and extracts cross-section elevation change data from the real-world 3D point cloud; simultaneously, it extracts texture change information of the cross-section region based on the pixel values ​​of UAV orthophoto images and oblique photogrammetry images, and constructs a grayscale sequence of candidate boundary cross-sections. Further, this module performs feature encoding based on the elevation undulation changes, continuity features, and texture perturbation degree in the grayscale sequence, generating a set of cross-section perturbation coded candidate boundaries. Subsequently, the planning anchoring constraint marking module calls the planning field boundary vector coordinates, permanent irrigation and drainage ditch design vector coordinates, and field road design vector coordinates in the storage unit to perform spatial matching analysis on the cross-sectional disturbance coding candidate boundary set. When the candidate boundary has spatial overlap, proximity connection, or topological association with the planning elements, the planning anchoring constraint marking is added to the corresponding candidate boundary to obtain the candidate boundary set with completed planning anchoring constraint marking, thereby distinguishing between stable boundaries with long-term planning constraints and temporary boundaries that may be generated by the construction process.

[0096] The system achieves phased construction boundary identification and stable plot boundary generation through construction trace removal and boundary reconstruction modules. Specifically, the controller calls the candidate boundary set with completed planning anchoring constraint markings and combines the cross-sectional disturbance coding attributes and planning anchoring marks of each candidate boundary to determine its stability. When a candidate boundary exhibits significant local elevation changes, identifies scraper marks, and lacks planning anchoring constraints, it is determined to be a phased construction trace boundary formed by land leveling, mechanical operations, temporary soil piling, etc., and is removed. Candidate boundaries with stable spatial disturbance characteristics and meeting planning anchoring constraint conditions are retained, thereby generating a set of stable plot boundary redrawing lines excluding phased construction traces. The final result is stored in a storage unit for use in land consolidation acceptance, plot management, and subsequent planning adjustments. Through the above system processing, intelligent redrawing of land consolidation plot boundaries based on UAV real-scene 3D mapping data, spatial disturbance feature analysis, and planning constraint fusion is realized, improving the accuracy of stable plot boundary extraction in complex construction environments.

Claims

1. A method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation, characterized in that, include: S1, acquire the pixel values ​​of UAV oblique photogrammetry images, the pixel values ​​of UAV orthophoto images, the 3D coordinate values ​​of real-world 3D point cloud, the vector coordinates of planned field boundaries, and the pre-configured thresholds, and generate a set of candidate boundaries for linear elevation textures that have completed spatial registration; S2, construct the candidate boundary cross-section elevation sequence and candidate boundary cross-section gray sequence along the candidate boundary set of the linear elevation texture that has been spatially registered, and generate the cross-section perturbation encoded candidate boundary set; S3, based on the planned field boundary vector coordinates, the permanent irrigation and drainage ditch design vector coordinates and the field road design vector coordinates, the cross-sectional disturbance coding candidate boundary set is marked with planning anchoring constraints to obtain the candidate boundary set with completed planning anchoring constraint marking; S4. Based on the cross-sectional disturbance coding attributes and candidate boundary planning anchor marks in the candidate boundary set that have completed the planning anchor constraint marking, the boundary with stage construction traces is eliminated, and a set of stable plot boundary redrawing lines that excludes stage construction traces is reconstructed and generated.

2. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation according to claim 1, characterized in that, In step S1, based on the pixel values ​​of the UAV oblique photographic image, the pixel values ​​of the UAV orthophoto image, the 3D coordinate values ​​of the real-world 3D point cloud, the vector coordinates of the planned field boundary, and a pre-configured threshold, a set of candidate boundaries for linear elevation textures that have completed spatial registration is generated, including: S11, the pixel values ​​of the UAV oblique photography image, the pixel values ​​of the UAV orthophoto image, and the three-dimensional coordinate values ​​of the real scene three-dimensional point cloud are uniformly projected to the coordinate system where the boundary vector coordinates of the planned field are located, forming a spatially consistent orthophoto point cloud grid. S12, extract point cloud elevation step line segments and orthophoto texture edge line segments from spatially consistent orthophoto point cloud raster to form elevation texture line segment pairs; S13. Match elevation texture line segment pairs according to the principle of minimum distance between line segment endpoints, and remove elevation texture line segment pairs whose length is less than the pre-configured minimum line segment length threshold to generate a set of candidate boundaries for linear elevation textures that have completed spatial registration.

3. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land boundaries for land consolidation according to claim 2, characterized in that, In step S11, the pixel values ​​of the UAV oblique photographic image, the pixel values ​​of the UAV orthophoto image, and the 3D coordinate values ​​of the real-world 3D point cloud are uniformly projected onto the coordinate system where the boundary vector coordinates of the planned field are located, forming a spatially consistent orthophoto point cloud raster, including: S111, based on the camera intrinsic parameter matrix, performs the back projection conversion of the pixel values ​​of the UAV oblique photographic image and the pixel values ​​of the UAV orthophoto image from the pixel coordinate system to the camera coordinate system. S112, based on the camera extrinsic parameter matrix, completes the rigid body transformation from the camera coordinate system to the surveying engineering coordinate system; S113, Complete the coordinate transformation from the surveying engineering coordinate system to the coordinate system where the boundary vector coordinates of the planned field are located, according to the coordinate system where the boundary vector coordinates of the planned field are located; S114. Establish a spatial index grid covering the vector coordinate range of the boundary of the planned field plots according to a unified raster resolution, and form a spatially consistent orthophoto point cloud raster containing raster elevation values, raster gray values ​​and spatial correspondence of image point clouds.

4. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation according to claim 2, characterized in that, In step S12, point cloud elevation step segments and orthophoto texture edge segments are extracted from the spatially consistent orthophoto point cloud raster to form elevation texture segment pairs, including: S121, calculate the horizontal elevation difference and vertical elevation difference according to the numerical difference of the raster elevation values ​​of the spatial index grid cells, and connect the edges of the spatial index grid cells that reach the pre-configured elevation change threshold to form point cloud elevation step segments. S122, calculate the gray-level gradient magnitude and gray-level gradient direction according to the gray-level difference between adjacent spatial index grid cells, and connect the edges of spatial index grid cells that reach the pre-configured gray-level gradient threshold to form orthophoto texture edge segments; S123, establish candidate pairing relationships for line segments based on the spatial proximity between point cloud elevation step line segments and orthophoto texture edge line segments; S124, write the point cloud elevation step line segments that are spatially adjacent and whose main direction angles meet the pre-configured direction consistency conditions and the orthophoto texture edge line segments into the same elevation texture line segment pair.

5. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation according to claim 1, characterized in that, In step S2, the candidate boundary cross-section elevation sequence and candidate boundary cross-section grayscale sequence are constructed along the candidate boundary set of the spatially registered linear elevation texture, including: S21. Along the tangent direction of each candidate boundary in the set of candidate boundaries of the linear elevation texture that has been spatially registered, generate a sequence of point cloud cross-sections perpendicular to the candidate boundaries according to a fixed sampling interval, forming a sequence of candidate boundary cross-section elevations. S22, calculate the center height difference, the slope sign on both sides, the bottom width of the trench and the mean difference of gray values ​​on both sides based on the candidate boundary cross section high program sequence and the candidate boundary cross section gray value sequence, and determine the cross section disturbance coding attributes; S23, write the cross-sectional perturbation coding attribute into the corresponding candidate boundary attribute in the linear elevation texture candidate boundary set that has completed spatial registration, and generate the cross-sectional perturbation coding candidate boundary set.

6. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation according to claim 5, characterized in that, In step S22, the center elevation difference, the slope sign on both sides, the bottom width of the trench, and the difference in the mean gray value on both sides are calculated based on the candidate boundary cross-section height sequence and the candidate boundary cross-section gray-scale sequence, and the cross-section perturbation coding attributes are determined, including: S221, calculate the scraper edge determination result according to the single-sided continuous slope break rule, and encode the candidate boundary that satisfies the scraper edge determination result as the scraper edge perturbation code; S222, calculate the shallow ditch edge determination result according to the rule of double-sided slope folding and holding the bottom of the depression, and encode the candidate boundary that satisfies the shallow ditch edge determination result as the shallow ditch edge perturbation code; S223, determine the non-construction disturbance determination result based on the shovel mark edge determination result and the shallow trench edge determination result, and encode the candidate boundary that satisfies the non-construction disturbance determination result as the non-construction disturbance code.

7. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation according to claim 1, characterized in that, In step S3, the candidate boundary set for cross-sectional disturbance coding is marked with planned anchorage constraints to obtain a candidate boundary set with completed planned anchorage constraint marking, including: S31, each candidate boundary in the cross-sectional disturbance coding candidate boundary set is matched with the nearest distance vector coordinates of the planned field boundary, the design vector coordinates of the permanent irrigation and drainage ditch, and the design vector coordinates of the field road. The endpoint anchorage distance from the endpoint of the candidate boundary to the planned node and the ratio of the planned fitting length between the candidate boundary and the planned line segment are calculated. S32, Generate candidate boundary planning anchor marks according to the rule that the endpoint anchorage distance is not greater than the pre-configured endpoint anchorage distance threshold and the planned fitting length ratio is not less than the pre-configured planned fitting ratio threshold. S33, according to the candidate boundary planning anchorage mark and the connectivity relationship of adjacent candidate boundary endpoints, update the cross-section disturbance coding candidate boundary set to the candidate boundary set that has completed the planning anchorage constraint mark.

8. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation according to claim 1, characterized in that, In step S4, based on the cross-sectional disturbance coding attributes and the planned anchorage marks of the candidate boundaries in the completed planning anchorage constraint mark set, the boundaries with staged construction traces are eliminated, and a set of stable plot boundary redrawing lines excluding staged construction traces is reconstructed and generated, including: S41, Based on the candidate boundary set with cross-sectional disturbance coding attribute of scraper edge disturbance code and candidate boundary planning anchorage mark as unanchored in the candidate boundary set with completed planning anchorage constraint mark, determine the scraper edge removal line segment set; S42, Based on the candidate boundary set with cross-sectional disturbance coding attribute of shallow ditch edge disturbance code and candidate boundary planning anchorage mark of unanchored state in the candidate boundary set with completed planning anchorage constraint mark, determine the set of shallow ditch edge elimination line segments; S43, delete the set of scraper edge removal line segments and the set of shallow ditch edge removal line segments from the candidate boundary set that has completed the planning anchoring constraint marking, and reconstruct the closed plot boundary according to the connection relationship of the endpoints of the candidate boundary, and generate a set of stable plot boundary redrawing lines that excludes stage construction traces.

9. The method for UAV-based real-scene 3D mapping and intelligent redrawing of land parcel boundaries for land consolidation according to claim 8, characterized in that, In step S43, the set of scraper edge removal segments and the set of shallow ditch edge removal segments are deleted from the candidate boundary set after the planning anchoring constraint marking is completed. Then, the closed plot boundary is reconstructed according to the connectivity of the retained candidate boundary endpoints, generating a set of stable plot boundary redrawing lines that excludes staged construction traces, including: S431, merge the candidate boundary unique identifiers in the set of scraper edge removal segments and the candidate boundary unique identifiers in the set of shallow trench edge removal segments to form a construction trace removal identifier index; S432, delete the candidate boundaries that hit the construction trace removal identifier index from the candidate boundary set that has completed the planning anchoring constraint mark, and construct the remaining candidate boundary connectivity graph based on the candidate boundary node coordinates and the connectivity relationship of the adjacent candidate boundary endpoints of the candidate boundaries that have not been deleted; S433, merge endpoints with the same name according to the coincidence of endpoint coordinates, and trace closed loops according to the line segment topology; S434. Based on the self-intersection check results and the area zero value check results, determine the effective closed loops and write the effective closed loops into the set of stable plot boundary redrawing lines that excludes stage construction traces.

10. A system for real-scene 3D mapping and intelligent redrawing of land parcel boundaries using unmanned aerial vehicles (UAVs) in land consolidation, characterized in that, include: The module includes data registration and candidate boundary generation, cross-section disturbance coding, planning anchorage constraint marking, construction trace removal and boundary reconstruction, storage unit, and controller. The storage unit is used to store the pixel values ​​of UAV oblique photogrammetry images, the pixel values ​​of UAV orthophoto images, the 3D coordinate values ​​of real-scene 3D point cloud, the vector coordinates of planned field boundaries, the design vector coordinates of permanent irrigation and drainage ditches, the design vector coordinates of field roads, pre-configured thresholds, a set of candidate boundaries for linear elevation textures that have completed spatial registration, a set of candidate boundaries for cross-section disturbance coding, a set of candidate boundaries for planned anchoring constraint marking, and a set of stable plot boundary redrawing lines that excludes staged construction traces; the controller is connected to the data registration and candidate boundary generation module, the cross-section disturbance coding module, the planned anchoring constraint marking module, the construction trace removal and boundary reconstruction module, and the storage unit, and is configured to execute the UAV real-scene 3D mapping and intelligent plot boundary redrawing method for land consolidation as described in any one of claims 1 to 9.

Images