Abnormal use area pattern generation method and device based on LiDAR point cloud

Abstract

The application relates to the technical field of point cloud modeling, in particular to an abnormal land use area graph generation method and device based on LiDAR point cloud, which comprises the following steps: acquiring LiDAR point cloud data and image data of a target area; performing point cloud segmentation processing on the pretreated LiDAR point cloud data to obtain a plurality of point cloud clusters; determining the spatial distribution characteristics of the point cloud clusters according to the point cloud curvature and coordinates of each sampling point in the point cloud clusters; determining an adaptive neighborhood value according to the spatial distribution characteristics of the point cloud clusters; determining a plurality of local normal vectors of the point cloud clusters based on the adaptive neighborhood value; determining the point cloud clusters corresponding to illegal buildings according to the direction cosine distribution characteristics of the plurality of local normal vectors of each point cloud cluster; and generating an abnormal land use area graph according to the point cloud clusters corresponding to the illegal buildings in the target area and the image data. The application improves the identification accuracy of illegal buildings in farmland.

Description

Technical Field

[0001] This invention relates to the field of point cloud modeling technology, and specifically to a method and apparatus for generating abnormal land use area graphics based on LiDAR point clouds. Background Technology

[0002] With the increasing demand for land resource supervision, remote sensing technology has become an important means of surface monitoring. LiDAR technology can actively acquire high-precision three-dimensional point cloud data of the land surface, providing an effective data foundation for land use analysis. Especially in densely vegetated areas, its penetration ability is superior to optical remote sensing, thus showing application potential in areas such as illegal building inspection.

[0003] Currently, most LiDAR point cloud-based building recognition methods rely on elevation changes, reflection intensity, or simple shape features for detection. These methods have significant limitations in complex farmland environments. Because vegetation and buildings in farmland scenes are highly similar in point cloud morphology, density, and distribution, existing technologies struggle to accurately distinguish between crops, trees, and illegal buildings, leading to numerous false positives and false negatives in the recognition results. This fails to meet the accuracy and completeness requirements of automated recognition for precise regulatory oversight. Summary of the Invention

[0004] To address the technical problem of inaccurate point cloud feature extraction for illegal buildings in farmland environments, leading to incomplete modeling and high false positive rates, this invention aims to provide a method and apparatus for generating abnormal land use area graphics based on LiDAR point clouds. The specific technical solution adopted is as follows: Firstly, a method for generating anomalous land use area graphics based on LiDAR point clouds is provided. This method includes: acquiring LiDAR point cloud data and image data of the target area, and preprocessing the LiDAR point cloud data; performing point cloud segmentation on the preprocessed LiDAR point cloud data to obtain multiple point cloud clusters, and determining the spatial distribution characteristics of the point cloud clusters based on the point cloud curvature and coordinates of each sampling point in the clusters, whereby the spatial distribution characteristics characterize the spatial sparsity of the point cloud clusters; for each point cloud cluster, determining an adaptive neighborhood value based on the spatial distribution characteristics of the clusters, and... Multiple local normal vectors of a point cloud cluster are determined based on adaptive neighborhood values. These local normal vectors represent the normal vectors of the plane formed by fitting the sampling point with its neighboring sampling points in the adaptive neighborhood values. Based on the distribution characteristics of the direction cosines of the multiple local normal vectors of each point cloud cluster, the point cloud cluster corresponding to the illegal building is determined. The distribution characteristics represent the frequency of each cosine value interval in the multiple cosine value intervals. The frequency is the number of direction cosines of the local normal vectors contained in the cosine value interval. Based on the point cloud clusters corresponding to the illegal buildings in the target area and the image data, the illegal buildings are modeled and textures are mapped to generate an abnormal land use area graphic.

[0005] In one possible design, the spatial distribution characteristics of the point cloud cluster are determined based on the point cloud curvature and coordinates of each sampling point in the point cloud cluster. This includes: determining the standard deviation and mean curvature of the point cloud based on the point cloud curvature of each sampling point in the point cloud cluster; determining the curvature distribution characteristics of the point cloud cluster based on the point cloud curvature, standard deviation, and mean curvature of all sampling points in the point cloud cluster; and determining the spatial distribution characteristics of the point cloud cluster based on the curvature distribution characteristics and the coordinates of each sampling point in the point cloud cluster.

[0006] In one possible design, the adaptive neighborhood value is determined based on the spatial distribution characteristics of the point cloud cluster, including: normalizing the spatial distribution characteristics of the point cloud cluster; and determining the adaptive neighborhood value based on the normalized spatial distribution characteristics of the point cloud cluster, the preset maximum neighborhood value, and the preset minimum neighborhood value.

[0007] In one possible design, multiple local normal vectors of a point cloud cluster are determined based on adaptive neighborhood values, including: for each sampling point in the point cloud cluster, determining a fitting plane between the sampling point and its neighboring sampling points based on the least squares method; and determining the normal vector of the fitting plane as the local normal vector corresponding to the sampling point, thereby obtaining multiple local normal vectors of the point cloud cluster.

[0008] In one possible design, the point cloud cluster corresponding to the illegal building is determined based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster. This includes: determining the direction cosine of the local normal vector corresponding to each sampling point in the point cloud cluster; determining the frequency of each cosine value interval; for each cosine value interval, if the frequency is greater than a first threshold, the cosine value interval is determined as a peak interval, where the first threshold is positively correlated with the number of sampling points included in the point cloud cluster; determining the number of peak intervals, and if the number of peak intervals is less than a second threshold, the point cloud cluster is determined as the point cloud cluster corresponding to the illegal building, where the second threshold is determined based on the symmetry characteristics of the point cloud cluster.

[0009] In one possible design, after determining the point cloud cluster corresponding to the illegal building, the method further includes: determining the extreme values ​​of the horizontal and vertical coordinates of the sampling points in the point cloud cluster corresponding to the illegal building; determining the illegal building area based on the extreme values ​​of the horizontal and vertical coordinates, wherein the illegal building area is a spatial area formed by expanding a preset range based on the extreme values ​​of the horizontal and vertical coordinates; and determining all sampling points located within the illegal building area as the point cloud cluster corresponding to the illegal building.

[0010] In one possible design, based on the point cloud clusters and image data corresponding to illegal buildings within the target area, the illegal buildings are modeled and textured to generate anomaly land use area graphics. This includes: calculating the transformation matrix between the point cloud clusters corresponding to the illegal buildings and the image data using a feature matching algorithm, and registering the point cloud clusters corresponding to the illegal buildings; reconstructing the registered point cloud clusters corresponding to the illegal buildings to generate a triangular mesh model; performing texture mapping on the triangular mesh model based on the image data to obtain a textured illegal building model; and overlaying the illegal building model onto a satellite map to generate anomaly land use area graphics.

[0011] In one possible design, the LiDAR point cloud data is preprocessed, including: using a preset filtering algorithm to process the LiDAR point cloud data, identifying and removing outliers, which are sampling points whose spatial distance from other sampling points exceeds a preset distance.

[0012] In one possible design, the preprocessed LiDAR point cloud data is segmented to obtain multiple point cloud clusters, including: determining the point cloud curvature of each sampling point in the preprocessed LiDAR point cloud data; and using a region growing segmentation algorithm to divide the preprocessed LiDAR point cloud data into multiple point cloud clusters based on the point cloud curvature of the sampling points.

[0013] Secondly, a device for generating abnormal land use area graphics based on LiDAR point clouds is provided, comprising: a data acquisition unit for acquiring LiDAR point cloud data and image data of a target area, and preprocessing the LiDAR point cloud data; a point cloud segmentation unit for segmenting the preprocessed LiDAR point cloud data to obtain multiple point cloud clusters; a feature analysis unit for determining the spatial distribution characteristics of the point cloud clusters, which characterize the spatial sparsity of the point cloud clusters; a normal vector calculation unit for determining an adaptive neighborhood value for each point cloud cluster based on its spatial distribution characteristics, and determining multiple local normal vectors of the point cloud cluster based on the adaptive neighborhood value, whereby the local normal vectors characterize the normal vectors of the plane formed by fitting the sampling point with the adaptive neighborhood value and its neighboring sampling points; and a cluster identification unit for identifying the point cloud clusters corresponding to illegal buildings based on the distribution characteristics of the direction cosines of the multiple local normal vectors of each point cloud cluster, whereby the distribution characteristics characterize the frequency of each cosine value interval, and the frequency is the number of direction cosines of the local normal vectors contained within the cosine value interval. The graphics generation unit is used to model and map textures of illegal buildings based on point cloud clusters and image data corresponding to illegal buildings within the target area, and generate graphics of abnormal land use areas.

[0014] The present invention has the following beneficial effects: The method for generating abnormal land use area graphics based on LiDAR point clouds provided by this invention can effectively overcome the misjudgment problem caused by the similarity of vegetation and building features in complex farmland environments. It significantly improves the pertinence and accuracy of geometric feature analysis of point cloud clusters through spatial distribution features and adaptive neighborhood mechanisms. It uses directional cosine distribution features to quantitatively evaluate the regularity of building surfaces, ensuring the complete extraction and reliable identification of illegal building point cloud clusters. At the same time, it generates textured illegal building models through multi-source data fusion and 3D reconstruction technology and integrates them into the geospatial background. Finally, it forms an abnormal land use area graphic that intuitively shows the relationship between illegal buildings and their surrounding locations, greatly improving the automation level and decision-making efficiency of land supervision. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a method for generating anomalous land use area graphics based on LiDAR point clouds, provided as an embodiment of the present invention; Figure 2 A flowchart illustrating a method for generating anomalous land use area graphics based on LiDAR point clouds, provided as an embodiment of the present invention; Figure 3 A flowchart illustrating a method for generating anomalous land use area graphics based on LiDAR point clouds, provided as an embodiment of the present invention; Figure 4 This invention provides a histogram of direction cosine values ​​and frequencies based on a vegetation-corresponding point cloud cluster, as an embodiment of the present invention. Figure 5 This invention provides a histogram of the direction cosine value and frequency of a point cloud cluster corresponding to a building. Figure 6 A flowchart illustrating a method for generating anomalous land use area graphics based on LiDAR point clouds, provided as an embodiment of the present invention; Figure 7 This is a schematic diagram of a device for generating abnormal land use area graphics based on LiDAR point clouds, provided in an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method and apparatus for generating abnormal land use area graphics based on LiDAR point clouds according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] In embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0019] In the description of this invention, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" and "more than one" refer to two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details the specific scheme of the method and apparatus for generating abnormal land use area graphics based on LiDAR point clouds provided by the present invention.

[0022] Please see Figure 1 The diagram illustrates a flowchart of a method for generating abnormal land use area graphics based on LiDAR point clouds according to an embodiment of the present invention, including steps S101-S105.

[0023] S101. Acquire LiDAR point cloud data and image data of the target area, and preprocess the LiDAR point cloud data.

[0024] One possible approach is to simultaneously scan and photograph the target area using a Light Detection and Ranging (LiDAR) sensor and a high-definition camera mounted on a drone, thereby acquiring three-dimensional LiDAR point cloud data and image data within the target area. Furthermore, the acquired LiDAR point cloud data within the target area is preprocessed.

[0025] In some embodiments, the LiDAR point cloud data is preprocessed by using a preset filtering algorithm to identify and remove outliers.

[0026] Outliers are sampling points whose spatial distance from other sampling points exceeds a preset distance.

[0027] It should be noted that since point cloud data such as buildings and vegetation are arranged in dense and regular blocks with small neighborhood distances between sampling points, while birds appear as scattered and isolated points, and birds are dynamic targets that may only appear at a certain moment during the LiDAR sensor scanning process, the number of point clouds is extremely small and there is no continuous spatial distribution pattern. Therefore, the preset distance can be set to 0.1 meters, 0.15 meters, etc. based on experience. This embodiment of the invention does not make specific limitations on this.

[0028] In some embodiments, the preset filtering algorithm can be an outlier filtering algorithm, with the number of neighboring points set to 15 and the standard layer multiple threshold set to 1.5. The collected LiDAR point cloud data is then input into the outlier filtering algorithm, and the point cloud data after removing outliers is output.

[0029] In some embodiments, LiDAR point cloud data can be preprocessed using a radius-based outlier filtering algorithm or by voxel grid downsampling for data simplification and smoothing to remove outliers and noise from the LiDAR point cloud data while retaining the point cloud data of buildings and vegetation.

[0030] S102. Perform point cloud segmentation on the preprocessed LiDAR point cloud data to obtain multiple point cloud clusters.

[0031] As one possible implementation, the point cloud curvature of each sampling point in the preprocessed LiDAR point cloud data is first calculated using the open-source point cloud data processing library PCL (Point Cloud Library). Then, a preset clustering algorithm is used to group sampling points with point cloud curvature within a preset similar range into the same group to form a point cloud cluster, thereby obtaining multiple point cloud clusters in the LiDAR point cloud data.

[0032] It should be noted that there is a significant difference in curvature between the point cloud data of vegetation (crops, trees, weeds, etc.) and buildings in farmland. Vegetation is characterized by irregular curvature with large fluctuations and relatively large values, while buildings are mostly planar with regular angles, resulting in more stable and concentrated point cloud curvature fluctuations. Therefore, point clouds with similar characteristics can be segmented into multiple clusters based on their curvature. The preset similarity range is a threshold interval set based on the point cloud curvature distribution patterns of common objects within the target area. This range is used to define the criteria for determining "similar curvature," ensuring that sampling points corresponding to the same object are classified into the same cluster, while sampling points corresponding to different objects are classified into different clusters.

[0033] In some embodiments, after determining the point cloud curvature of each sampling point in the preprocessed LiDAR point cloud data, a region growing segmentation algorithm can be used to divide the preprocessed LiDAR point cloud data into multiple point cloud clusters based on the point cloud curvature of the sampling points.

[0034] In this algorithm, the growth threshold is set to an empirical value of 0.5 and the neighborhood search radius is set to 8. After inputting the curvature of the point cloud at the sampling points, the point cloud data is divided into multiple point cloud clusters corresponding to different land features based on the curvature characteristics.

[0035] In some embodiments, point cloud clusters can also be divided based on a multi-feature fusion segmentation method. In this method, the point cloud curvature and point cloud reflection intensity of the sampling points are obtained as joint segmentation features. By using preset fusion rules, such as weighted calculation of feature similarity, the correlation of the sampling points is determined. Sampling points that satisfy the similarity conditions of both types of features are divided into the same cluster, thereby improving the segmentation accuracy through multi-feature complementarity.

[0036] S103. Determine the spatial distribution characteristics of the point cloud cluster based on the point cloud curvature and coordinates of each sampling point in the point cloud cluster.

[0037] Among them, spatial distribution characteristics are used to characterize the spatial sparsity of point cloud clusters.

[0038] As one possible implementation, based on the point cloud curvature of each sampling point in the point cloud cluster, point cloud curvature-related parameters are calculated, including the standard deviation of curvature and the mean curvature. The standard deviation of curvature is used to characterize the dispersion of point cloud curvature within the point cloud cluster. A larger value indicates a more dispersed distribution of point cloud curvature, while a smaller value indicates a more concentrated distribution of point cloud curvature. The mean curvature is used to characterize the average level of point cloud curvature within the point cloud cluster, reflecting the overall bending trend of the cluster. By further combining point cloud curvature, standard deviation of curvature, and mean curvature, the curvature distribution characteristics can be determined.

[0039] Furthermore, based on the coordinates of each sampling point, the spatial distance between adjacent sampling points is calculated. This spatial distance characterizes the proximity of sampling points in space; a smaller distance indicates a denser sampling point cluster, while a larger distance indicates a sparser one. Further, the curvature distribution characteristics and the spatial distance between adjacent sampling points are fused to determine the spatial distribution characteristics used to determine the spatial sparsity of the standard point cloud cluster. If the point cloud curvature distribution is uniform and the spatial distance between adjacent sampling points is small, the smaller the value of the spatial distribution characteristic, indicating a denser sampling point cluster. Conversely, if the point cloud curvature distribution is dispersed and the spatial distance between adjacent sampling points is large, the larger the value of the spatial distribution characteristic, indicating a sparser sampling point cluster.

[0040] It should be noted that the specific implementation method for determining the spatial distribution characteristics of the point cloud cluster based on the point cloud curvature and coordinates of each sampling point in the point cloud cluster can be referred to the description of steps S201-S203 in the subsequent embodiments of the present invention, and will not be repeated here.

[0041] In some embodiments, to determine the spatial distribution characteristics of a point cloud cluster, after calculating the curvature distribution characteristics based on the point cloud curvature of each sampling point in the point cloud cluster, the number of sampling points in each unit space is counted based on the coordinates of each sampling point, and the average number of sampling points per unit space is determined to obtain the point cloud density. The curvature distribution characteristics and point cloud density are further fused, such as through weighted calculation, to obtain the spatial distribution characteristics. Specifically, if the point cloud curvature distribution is uniform and the point cloud density is high, the value of the spatial distribution characteristics is small, and the sparsity of the point cloud in the cluster is low. Conversely, if the point cloud curvature distribution is dispersed and the point cloud density is low, the value of the spatial distribution characteristics is large, and the sparsity of the point cloud in the cluster is high.

[0042] In some embodiments, to determine the spatial distribution characteristics of a point cloud cluster, after determining the curvature distribution characteristics and spatial distance between adjacent sampling points, the consistency of normal vectors (such as normal vector deviation values) can be calculated based on the point cloud normal vectors of each sampling point. This further integrates the three types of features, such as through weighted calculation, to obtain the spatial distribution characteristics. Specifically, if the point cloud curvature distribution is uniform, the spatial distance is small, and the normal vector consistency is high, the value of the spatial distribution characteristic is small, indicating low point cloud sparsity within the cluster. Conversely, if the point cloud curvature distribution is scattered, the spatial distance is large, and the normal vector consistency is low, the value of the spatial distribution characteristic is large, indicating high point cloud sparsity within the cluster. This allows for the determination of the sparsity of the point cloud cluster based on more features, improving accuracy.

[0043] S104. For each point cloud cluster, determine the adaptive neighborhood value based on the spatial distribution characteristics of the point cloud cluster.

[0044] As one possible approach, firstly, based on the spatial distribution characteristics of multiple point cloud clusters, normalization processing is performed to transform the spatial distribution characteristics of different point cloud clusters into a unified numerical range, eliminating the problem of inconsistent feature value scales caused by differences in point cloud cluster size and sampling density, and ensuring the comparability of adaptive neighborhood value calculation.

[0045] Optionally, for the spatial distribution characteristics of multiple point cloud clusters, normalization processing can be performed. This can be done using methods such as deviation standardization, maximum and minimum value normalization, maximum value normalization, and Z-score.

[0046] Furthermore, for each point cloud cluster, an adaptive neighborhood value is calculated and rounded up based on the normalized spatial distribution characteristics of the cluster and the initial neighborhood value. This adaptive neighborhood value reflects the physical meaning of the number of sampling points within the neighborhood. The adaptive neighborhood value is positively correlated with the spatial distribution characteristics. A larger spatial distribution characteristic value indicates a sparser point cloud cluster, requiring a larger neighborhood value to include enough sampling points to capture the morphological features of the point cloud within the neighborhood. Conversely, a smaller spatial distribution characteristic value indicates a denser point cloud cluster, requiring a smaller neighborhood value to capture the detailed features of the point cloud.

[0047] It should be noted that, due to the irregularity of the vegetation surface, the curvature dispersion of local sampling points within its point cloud cluster is relatively large, and the proportion of sparse point clouds within the cluster is relatively large. A larger adaptive neighborhood value can ensure that enough neighboring sampling points are selected to fully reflect the local surface morphology. On the other hand, the curvature dispersion of relatively regular building point cloud clusters is smaller, and the proportion of dense point clouds is larger. A smaller adaptive neighborhood value can avoid mixing in multi-planar point clouds and accurately fit the normal vector of a single surface.

[0048] In some embodiments, after normalizing the spatial distribution characteristics of the point cloud cluster, a preset maximum and a preset minimum neighborhood value are introduced to limit the adaptive neighborhood value to the range of the preset minimum and the preset maximum neighborhood value. This avoids the adaptive neighborhood value being too large due to excessively sparse sampling points in the point cloud cluster, which could mix in the point cloud data of other objects, such as extending the neighborhood of vegetation point cloud to distant buildings. It also avoids the adaptive neighborhood value being too small due to excessively dense sampling points in the point cloud cluster, which could not cover enough neighboring sampling points, such as insufficient neighborhood points at small protrusions on walls, making it difficult to fit an accurate plane.

[0049] In some embodiments, based on the spatial distribution characteristics of the point cloud cluster after normalization, the preset maximum value of the neighborhood, and the preset minimum value of the neighborhood, the following formula is used to determine the adaptive neighborhood value: In the formula, Indicates the first Adaptive neighborhood values ​​for a point cloud cluster This represents the function that takes the minimum value. This indicates rounding up to ensure adaptive neighborhood values. It is a positive integer. This represents the function that takes the maximum value. This represents the initial neighborhood value, set to an empirical value of 10. Indicates the first Spatial distribution characteristics of point cloud clusters This represents the spatial distribution characteristics after normalization. The hyperparameters are preset and are extremely small positive numbers. This invention does not impose any special restrictions on them; in this embodiment, they are... For example, its purpose is to avoid the denominator being 0 during fraction operations. This represents the preset minimum neighborhood value, which is empirically set to 5. This represents the preset maximum neighborhood value, which is empirically set to 30. This indicates taking the absolute value.

[0050] In this system, the adaptive neighborhood value of each point cloud cluster is dynamically updated as the spatial distribution characteristics change; if the spatial distribution characteristics increase, the adaptive neighborhood value increases. For the th... A point cloud cluster, if its spatial distribution characteristics The larger the value, the greater the coefficient for adjusting the initial neighborhood value. The larger the value, the more likely it is to be adjusted after adjusting the initial neighborhood value. The larger the value, the more adaptive the neighborhood value will be within the range of the preset minimum neighborhood value to the preset maximum neighborhood value. The larger the value, the better, ensuring that enough neighboring sampling points are selected to fully reflect the local surface morphology; conversely, if its spatial distribution characteristics... The smaller the value, the smaller the coefficient for adjusting the initial neighborhood value. The smaller the value, the better the value obtained after adjusting the initial neighborhood values. The smaller the value, the more adaptive the neighborhood value will be within the range of the preset minimum neighborhood value to the preset maximum neighborhood value. The smaller the value, the better, avoiding the mixing of multi-plane point clouds and accurately fitting the normal vector of a single surface.

[0051] Understandably, point cloud clusters corresponding to vegetation in farmland have sparse sampling points, while point cloud clusters corresponding to buildings have dense sampling points. If a fixed neighborhood value is used to divide the neighborhood and calculate the normal vector of the sampling points, it will lead to a large deviation in the normal vector. For example, for dense point cloud clusters of buildings, if the neighborhood value is too small, small protrusions or depressions on the wall surface may cause the normal vector to tilt locally, resulting in large fluctuations. If the neighborhood value is too large, the point set may contain mixed information from multiple planes (such as wall corners and the ground, wall sides and roofs), which will also lead to a significant deviation in the subsequent calculation of the normal vector of the neighborhood. This invention determines adaptive neighborhood values ​​related to the sparsity of the point cloud cluster by utilizing the spatial distribution characteristics of the point cloud cluster, thus enabling more accurate calculation of the sampling point normal vector based on the neighborhood divided by the adaptive neighborhood value.

[0052] S105. Determine multiple local normal vectors of a point cloud cluster based on adaptive neighborhood values.

[0053] Among them, the local normal vector is used to characterize the normal vector of the plane formed by fitting the sampling point with the adaptive neighborhood value and the neighboring sampling points.

[0054] As one possible implementation, for each sampling point in the point cloud cluster, based on the adaptive neighborhood value determined in step S104 above, a number of sampling points with adaptive neighborhood values ​​adjacent to the sampling point are selected from the point cloud cluster. Further, based on the sampling point and the selected number of adjacent sampling points with adaptive neighborhood values, a plane fitting is performed. The normal vector of the fitted plane is determined, and this normal vector is defined as the local normal vector corresponding to the sampling point. Each sampling point in the point cloud cluster is further processed to obtain multiple local normal vectors for the point cloud cluster.

[0055] Understandably, the fitted plane is an approximate representation of the real surface where the sampling point is located. The orientation (direction, degree of tilt) of the plane is similar to the local real surface, so the normal vector of the fitted plane is determined as the local normal vector corresponding to the sampling point, which can represent the orientation of the local surface where the sampling point is located.

[0056] In some embodiments, for each sampling point in the point cloud cluster, a fitting plane between the sampling point and its adaptive neighborhood values ​​of neighboring sampling points is determined based on the least squares method. The normal vector of the fitting plane is further determined as the local normal vector corresponding to the sampling point, thereby obtaining multiple local normal vectors representing the overall surface orientation of the point cloud cluster.

[0057] Alternatively, plane fitting algorithms such as Random Sample Consensus (RANSAC) and Moving Least Squares can be used to fit the sampling point to the adaptive neighborhood value and its neighboring sampling points into a plane.

[0058] In some embodiments, principal component analysis can be used to extract the three principal component directions of the point set (the sampling point and its adaptive neighborhood values ​​and neighboring sampling points). The third principal component direction, which is perpendicular to the first two principal components (reflecting the distribution of the point set in the plane), is the local normal vector corresponding to the sampling point. The principal component directions are determined based on the variance distribution of the point set coordinates. The first two principal components correspond to the plane directions where the point set distribution is most extensive, and the third principal component corresponds to the perpendicular direction of the plane.

[0059] S106. Based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster, determine the point cloud cluster corresponding to the illegal building.

[0060] Among them, the distribution feature is used to characterize the frequency of each cosine value interval in multiple cosine value intervals. The frequency is the number of direction cosines of the local normal vector contained in the cosine value interval.

[0061] As one possible implementation, for each point cloud cluster, multiple direction cosines of local normal vectors are determined, where the direction cosine is the cosine of the angle between the local normal vector and the coordinate axis. In this embodiment of the invention, since the Z-axis direction cosine of the local normal vector corresponding to the top sampling point of non-flat roof buildings (such as arched roofs and sloping roofs) fluctuates greatly, it is easy to cause feature confusion. However, the X-axis and Y-axis direction cosines can stably reflect the orientation consistency of regular surfaces such as walls. Therefore, in this embodiment of the invention, the direction cosine can be the X-axis direction cosine or the Y-axis direction cosine of the local normal vector.

[0062] Further, based on multiple preset cosine value intervals, cluster analysis is performed to determine the number of direction cosines of local normal vectors in each cosine value interval, forming a distribution feature that characterizes the surface orientation consistency of the point cloud cluster. When the direction cosines of local normal vectors are mostly concentrated in a few cosine value intervals, it indicates that the orientation consistency of the sampling points in the point cloud cluster is high, confirming that the point cloud cluster corresponds to an illegal building. When the direction cosines of local normal vectors are scattered across various cosine value intervals, it indicates that the orientation consistency of the sampling points in the point cloud cluster is low, confirming that the point cloud cluster corresponds to a non-building structure.

[0063] In some embodiments, the variance of the direction cosine can also be determined as a distribution feature. The variance of the direction cosine is used to characterize the dispersion of the direction cosine; the smaller the value, the more concentrated the direction cosine. Specifically, the variances of the X-axis and Y-axis direction cosines of the local normal vectors are calculated separately, and a variance threshold is set based on the historical dispersion level of the direction cosines of illegal buildings. If the variances of both the X-axis and Y-axis direction cosines are less than the variance threshold, it indicates that the direction consistency of the local normal vectors is high, and it is identified as a point cloud cluster corresponding to an illegal building. If either the variance of the X-axis or Y-axis direction cosine is greater than the variance threshold, it is identified as a point cloud cluster corresponding to a non-building.

[0064] It should be noted that the specific implementation method for determining the point cloud cluster corresponding to the illegal building based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster can be referred to the subsequent steps S201-S204 in the embodiment of the present invention, which will not be repeated here.

[0065] S107. Based on the point cloud clusters and image data corresponding to the illegal buildings in the target area, model and map the textures of the illegal buildings to generate abnormal land use area graphics.

[0066] One possible implementation involves first reconstructing a sparse 3D point cloud of the illegal building using Structure from Motion (SfM) technology based on image data. Then, a feature matching algorithm is used to calculate the transformation matrix between the point cloud cluster corresponding to the illegal building and the sparse 3D point cloud reconstructed from the image data. This is used to register the point cloud cluster corresponding to the illegal building, establishing an accurate coordinate correspondence for subsequent texture mapping. Further, the registered point cloud cluster corresponding to the illegal building is reconstructed to generate a triangular mesh model. Texture mapping is then performed on the triangular mesh model based on the image data to obtain a textured model of the illegal building. Finally, the illegal building model is overlaid on a satellite map to generate an abnormal land use area graphic for visualization, facilitating efficient investigation of abnormal land use by land inspectors in farmland.

[0067] In some embodiments, SfM technology is used to reconstruct the sparse three-dimensional point cloud of illegal buildings. For example, a high-definition camera mounted on a drone takes pictures of the target area during flight, collects multi-view images of the target area, extracts image feature points from the multi-view images and matches them, and calculates the sparse three-dimensional point cloud of illegal buildings in the target area using the multi-view geometry principle.

[0068] In some embodiments, the point cloud cluster corresponding to the illegal building can be calculated using the Intrinsic Shape Signature (ISS) feature matching algorithm and then registered with the transformation matrix of the sparse 3D point cloud reconstructed from the image data to obtain the registered point cloud cluster of the illegal building, thus improving the accuracy of the point cloud cluster of the illegal building. A Poisson reconstruction is then used to generate a building triangular mesh on the registered point cloud cluster of the illegal building, simplified to 100,000 faces (preserving corners, door and window boundaries). Finally, normal vector matching, homography transformation, parametric mapping, and multi-band fusion algorithms are used to perform texture mapping on the triangular mesh. A vertical view image is selected for each wall surface, perspective distortion is corrected, and the image is then pasted, with smooth transitions at the seams, resulting in a textured obj model of the illegal building. This model is then overlaid onto a satellite map to obtain a graphic of the abnormal land use area, visually demonstrating the positional relationship between the illegal building and farmland.

[0069] In some embodiments, a voxel model can be established based on the point cloud cluster of illegal buildings. The voxel model consists of several cubic units (voxels) in three-dimensional space. The existence of each voxel is determined by whether it contains sampling points of illegal buildings, which can intuitively represent the spatial occupancy pattern of illegal buildings. During texture mapping, the pixel information of the image data is assigned to the outer surface of the voxels based on the transformation matrix, resulting in a textured voxel model, which is then superimposed onto a satellite map to generate anomaly land use area graphics.

[0070] Understandably, the method for generating abnormal land use area graphics based on LiDAR point clouds provided by this invention can effectively overcome the misjudgment problem caused by the similarity of vegetation and building features in complex farmland environments. It significantly improves the pertinence and accuracy of geometric feature analysis of point cloud clusters through spatial distribution features and adaptive neighborhood mechanisms. It uses directional cosine distribution features to quantitatively evaluate the regularity of building surfaces, ensuring the complete extraction and reliable identification of illegal building point cloud clusters. At the same time, it generates textured illegal building models through multi-source data fusion and 3D reconstruction technology and integrates them into the geospatial background. Finally, it forms an abnormal land use area graphic that intuitively shows the relationship between illegal buildings and their surrounding locations, greatly improving the automation level and decision-making efficiency of land supervision.

[0071] In a design, such as Figure 2 As shown, the present invention provides a specific implementation method for determining the spatial distribution characteristics of a point cloud cluster based on the point cloud curvature and coordinates of each sampling point in the point cloud cluster, including the following steps S201-S203.

[0072] S201. Determine the standard deviation of curvature and the average curvature based on the curvature of each sampling point in the point cloud cluster.

[0073] S202. Determine the curvature distribution characteristics of the point cloud cluster based on the point cloud curvature, standard deviation of curvature, and average curvature of all sampling points in the point cloud cluster.

[0074] In some embodiments, the formula for calculating the curvature distribution characteristics is as follows: In the formula, Indicates the first Curvature distribution characteristics of a point cloud cluster Indicates the first The number of sampling points within a point cloud cluster Indicates the first The standard deviation of curvature of a point cloud cluster Indicates the first The average curvature of a point cloud cluster, This indicates that the absolute value is taken to ensure that the curvature distribution characteristic is positive. Indicates the first The first point cloud cluster The curvature of each sampling point , They represent the first The maximum and minimum curvature of the sampling points within a point cloud cluster. It is a very small positive number, for example, it can take the value 0.001. , , For preset weight values, ,For example , , The possible values ​​are 0.3, 0.4, and 0.3, respectively.

[0075] Among them, the standard deviation of curvature With curvature distribution characteristics Positive correlation, The larger the value, the higher the overall dispersion of the point cloud curvature within the point cloud cluster (e.g., irregular vegetation surfaces with fluctuating curvature), and the more pronounced the curvature distribution characteristics. It increases accordingly; conversely, The smaller (e.g., the more stable the curvature of the building plan). The smaller.

[0076] and The summation term shows a positive correlation, reflecting the curvature of each sampling point and the average curvature of the point cloud cluster. The sum of deviations avoids the problem of positive and negative deviations canceling each other out (e.g., some points have curvature above the mean while others are below the mean; standard deviation might mask details, while absolute deviation captures the total deviation completely). A larger sum indicates that the curvature of most sampling points within the cluster deviates significantly from the mean. The larger.

[0077] and The difference is positively correlated, quantifying the "range" of curvature within a cluster. Vegetation, due to its irregular surface, may have a curvature ranging from 0.2 (flat leaves) to 1.6 (sharp branches), a range of up to 1.4; while building floor plan curvature may only range from 0.3 to 0.5, a range of only 0.2. A larger difference indicates a more significant extreme difference in curvature distribution. The larger.

[0078] and The correlation is negative, and the core function of the denominator is "normalization," eliminating the influence of two irrelevant factors on the result. Interference. Among them, the sizes of different clusters vary greatly (e.g., a building cluster with 1000 points, a vegetation cluster with 200 points), if not divided by... The absolute bias and inherently larger bias of large-scale clusters will lead to It only reflects the "size" of the point cloud cluster, not its "dispersion", divided by Then, the absolute deviation can be converted into an "average deviation level" to ensure the accuracy of point cloud clusters of different sizes. Comparable; different clusters have different absolute values ​​of average curvature (e.g., average curvature of building plan is 0.4, average curvature of vegetation is 0.8), if not divided by... The absolute deviation of high mean curvature clusters may naturally be larger, leading to Confusing "absolute deviation" with "relative dispersion"; dividing by back, Focusing on the "relative degree of curvature deviation from the mean," for example, the absolute deviation of 0.2 is 50% for buildings with an average curvature of 0.4, but only 25% for vegetation with an average curvature of 0.8. Normalization can accurately reflect this difference.

[0079] It should be noted that the above calculation formula for determining the curvature distribution characteristics of point cloud clusters captures curvature distribution characteristics from three dimensions: standard deviation (overall fluctuation), normalized mean absolute deviation (point-level relative deviation), and range (global span). When distinguishing between farmland and building point clouds, vegetation surfaces not only exhibit large overall fluctuations but also often show local abrupt changes and wide-range variations. This multi-scale feature fusion can more sensitively and comprehensively amplify the differences in curvature irregularities between vegetation and buildings.

[0080] S203. Determine the spatial distribution characteristics of the point cloud cluster based on the curvature distribution characteristics and the coordinates of each sampling point in the point cloud cluster.

[0081] In some embodiments, the formula for calculating spatial distribution characteristics is as follows: In the formula, Indicates the first Spatial distribution characteristics of point cloud clusters Indicates the first Curvature distribution characteristics of a point cloud cluster Indicates the first The number of sampling points within a point cloud cluster Indicates the first Within the point cloud cluster, the first The coordinates of each sampling point Indicates the first Within the point cloud cluster, the first The coordinates of each sampling point This indicates the calculation of the Euclidean distance between two coordinates.

[0082] Among them, spatial distribution characteristics With curvature distribution characteristics There is a positive correlation: the greater the curvature distribution feature in a point cloud cluster, the sparser the sampling points, and the greater the spatial distribution feature; spatial distribution feature and They are positively correlated; as the average Euclidean distance increases, the sampling points become sparser and the spatial distribution characteristics become greater.

[0083] Understandably, in determining the spatial distribution characteristics of a point cloud cluster in this embodiment of the invention, the standard deviation and average value of curvature are first calculated based on the curvature of the point cloud. Then, a curvature distribution characteristic that can accurately characterize the uniformity of curvature distribution within the point cloud cluster is formed. This process elevates the discrete point cloud curvature into a reliable indicator describing the overall geometric regularity of the cluster. Subsequently, the curvature distribution characteristic is fused with the spatial distance information between sampling points to jointly determine the spatial distribution characteristics. This achieves an effective correlation between the geometric characteristics of the cluster surface and the spatial structural characteristics, so that the final obtained spatial distribution characteristics not only reflect the density of the point cloud but also contain deeper information about its structural regularity, providing a more comprehensive and stable feature input for the accurate calculation of subsequent adaptive neighborhood values.

[0084] In a design, such as Figure 3 As shown, the present invention provides a specific implementation method for determining the point cloud cluster corresponding to an illegal building based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster, including the following steps S301-S304.

[0085] S301. Determine the direction cosine of the local normal vector corresponding to each sampling point in the point cloud cluster.

[0086] As one possible implementation, the X-axis direction cosine and / or Y-axis direction cosine of the local normal vector corresponding to each sampling point are calculated.

[0087] S302. Determine the frequency of each cosine interval.

[0088] As one possible implementation, based on the range of cosine values, the system is divided into multiple cosine value intervals of equal length. Taking the cosine in the X-axis direction as an example, the number of X-axis cosine values ​​contained in each of the multiple cosine value intervals is determined as the frequency of the cosine value interval.

[0089] For example, the range of cosine values ​​is [-1, 1]. Taking an interval length of 0.2 as an example, the multiple cosine value intervals are [-1, -0.8), [-0.8, -0.6), [-0.6, -0.4), ... [0.4, 0.6), [0.6, 0.8), [0.8, 1]. Furthermore, the number of cosines in the X-axis direction within each cosine value interval is determined as the frequency of each cosine value interval.

[0090] S303. For each cosine value interval, if the frequency is greater than the first threshold, the cosine value interval is determined as the peak interval.

[0091] The first threshold is positively correlated with the number of sampling points included in the point cloud cluster.

[0092] As one possible implementation, for each cosine value interval, the frequency of the cosine value is compared with a first threshold. If the frequency is greater than the first threshold, the cosine value interval is determined to be the peak interval.

[0093] It should be noted that the first threshold can be set according to the total number of sampling points in the point cloud cluster. For example, the first threshold can be 10%, 15%, 20% of the total number, etc. This embodiment of the invention does not make specific limitations on this.

[0094] In some embodiments, a histogram can be created based on the frequency of each cosine interval to visually display the distribution of the local normal vectors corresponding to each sampling point in the point cloud cluster. For example, such as... Figure 4 As shown, this is a histogram of the direction cosine value and frequency based on the point cloud clusters corresponding to vegetation. Negative values ​​represent the cosine of the angle between the normal vector direction and the negative direction of the axis. Due to the irregularity and anisotropy of the vegetation surface in farmland, the normal vector directions in its point cloud clusters are relatively dispersed, resulting in a large number of cosine values ​​and a histogram with many dispersed peaks. The histogram of the cosine value interval and frequency for the point cloud clusters corresponding to buildings is shown below. Figure 5 As shown, since building walls are generally regular, symmetrical, and flat, the point cloud clusters are also relatively concentrated. The normal vector directions of the same wall are basically consistent, so the remaining chord values ​​are mainly concentrated around -1, 0, and 1, and the histogram peaks are fewer and more concentrated.

[0095] S304. Determine the number of peak intervals, and if the number of peak intervals is less than the second threshold, determine that the point cloud cluster is the point cloud cluster corresponding to the illegal building.

[0096] The second threshold is determined based on the symmetry characteristics of the point cloud cluster.

[0097] As one possible implementation, based on the peak intervals determined in step S303 above, if the number of peak intervals is less than the second threshold, the point cloud cluster is determined to be the point cloud cluster corresponding to the illegal building.

[0098] In some embodiments, the formula for calculating the second threshold is as follows: In the formula, This represents the second threshold. To preset the minimum second threshold, an empirical value of 2 can be taken. To preset the maximum second threshold, an empirical value of 5 can be taken. This represents the average error of a pair of symmetrical points. Indicates the maximum error. It is a very small positive number, for example, it can take the value 0.001. This is for rounding up.

[0099] in, The second threshold boundary is preset. The symmetry score is calculated based on symmetrical point pairs and ranges from 0 to 1. The better the symmetry of the point cloud cluster, the higher the value of the score, and the larger the final second threshold is. This makes it easier to meet the judgment condition that the number of peak intervals is less than the second threshold and identify the point cloud cluster corresponding to the illegal building.

[0100] In some embodiments, after determining the point cloud cluster corresponding to the illegal building, the sampling point label in the point cloud cluster corresponding to the illegal building is set to 1 to indicate that the sampling point is a sampling point of the illegal building, and the sampling point label in the point cloud cluster corresponding to the non-building is set to 0 to indicate that the sampling point is a sampling point of the non-building. In this way, the sampling point data used to generate the model can be determined according to the label of the sampling point in the subsequent model generation process.

[0101] Understandably, in the LiDAR point cloud-based method for generating abnormal land use area graphics provided by this invention, the surface orientation features of the point cloud cluster are first transformed into a quantifiable distribution model by statistically analyzing the distribution frequency of the direction cosines of all local normal vectors in different intervals. Then, a first threshold based on cluster size is introduced to objectively determine the peak intervals, effectively eliminating discrimination bias caused by differences in cluster size. Finally, a second threshold dynamically determined by combining cluster symmetry features is used to make the final decision on the number of peaks. This ensures that the identification process considers not only the concentration of surface orientation but also the essential characteristic of building structure—structural symmetry. This effectively distinguishes between randomly distributed vegetation clusters and building clusters with consistent surface orientation and regular structures. Especially when dealing with illegal buildings with complex roof structures, by ignoring the Z-axis direction cosine and focusing on the feature analysis of the walls in the X or Y directions, misjudgments caused by special roof shapes are avoided, significantly improving the accuracy and reliability of building identification results in complex real farmland environments.

[0102] In one design, since the roof and eaves of a building typically extend beyond the walls, to ensure the accuracy of modeling illegal structures, such as... Figure 6 As shown, the method for generating abnormal land use area graphics based on LiDAR point clouds provided by the present invention further includes the following steps S401-S403 after determining the point cloud cluster corresponding to the illegal building.

[0103] S401. Determine the extreme values ​​of the horizontal and vertical coordinates of the sampling points in the point cloud cluster corresponding to the illegal building.

[0104] As one possible implementation, based on the coordinates of all sampling points in the point cloud cluster corresponding to the illegal building, the maximum and minimum values ​​of the horizontal coordinate, as well as the maximum and minimum values ​​of the vertical coordinate, are determined.

[0105] S402. Determine the areas of illegal buildings based on the extreme values ​​of the horizontal and vertical axes.

[0106] Among them, the area of ​​illegal construction is a spatial area formed by expanding the preset range based on the extreme values ​​of the horizontal and vertical coordinates.

[0107] As one possible approach, based on the extreme values ​​of the horizontal coordinate and a preset range, the minimum and maximum values ​​of the horizontal coordinate of the illegal building area are determined. Based on the extreme values ​​of the vertical coordinate and a preset range, the minimum and maximum values ​​of the vertical coordinate of the illegal building area are determined. Then, based on the minimum, maximum, minimum, and maximum values ​​of the horizontal coordinate of the illegal building area, the spatial range of the illegal building area is obtained.

[0108] For example, the vertices of the spatial extent of the illegal building area can be represented as ( , ), ( , ), ( , ), ( , There are no restrictions on the Z-axis direction. These represent the minimum, maximum, minimum, and maximum values ​​of the x-coordinate, y-coordinate, and y-coordinate in the point cloud cluster corresponding to the illegal building. For the preset range, an empirical value of 0.2 can be used. These represent the minimum, maximum, minimum, and maximum values ​​of the horizontal axis of the area containing the illegal buildings.

[0109] S403. All sampling points located within the illegal building area are identified as the point cloud clusters corresponding to the illegal buildings.

[0110] As one possible approach, based on the minimum, maximum, minimum, and maximum values ​​of the horizontal coordinate of the illegal building area, all sampling points located within that area are selected, and the selected sampling points are identified as the point cloud cluster corresponding to the illegal building.

[0111] For example, the coordinates in the point cloud cluster corresponding to the illegal building satisfy the following inequality: In the formula, The minimum x-coordinate value of the area with illegal buildings. The x-coordinate of the area with illegal buildings is the maximum value. This represents the minimum value of the ordinate of the area with illegal buildings. The maximum value of the vertical coordinate of the area with illegal buildings. To expand the preset range, an empirical value of 0.2 can be used. , () represents the x and y coordinates of the sampling point.

[0112] In some embodiments, the label of all sampling points within the illegal building area is set to 1 to indicate that the sampling point is a sampling point of an illegal building. In this way, the sampling point data used to generate the model can be determined based on the label of the sampling point in the subsequent model generation process.

[0113] Understandably, in the method for generating abnormal land use area graphics based on LiDAR point clouds provided by this invention, after initially identifying the point cloud clusters of illegal building walls, a more inclusive illegal building area is constructed by extracting the extreme value range of its spatial coordinates and adding a preset floating value. This effectively solves the problem that the point clouds of the building top and the point clouds of the walls are misjudged as different clusters in the initial segmentation stage due to the overhanging eaves, irregular roofs, or uneven point cloud density, thus improving the integrity and accuracy of the reconstruction of the three-dimensional model of illegal buildings.

[0114] In one design, the present invention provides a device for generating images of anomalous land use areas based on LiDAR point clouds, such as... Figure 7 As shown, the LiDAR point cloud-based abnormal land use area graphic generation device 50 includes a data acquisition unit 501, a point cloud segmentation unit 502, a feature analysis unit 503, a normal vector calculation unit 504, a cluster identification unit 505, and a graphic generation unit 506.

[0115] The data acquisition unit 501 is used to acquire LiDAR point cloud data and image data of the target area, and to preprocess the LiDAR point cloud data. The point cloud segmentation unit 502 is used to perform point cloud segmentation processing on the preprocessed LiDAR point cloud data to obtain multiple point cloud clusters. The feature analysis unit 503 is used to determine the spatial distribution characteristics of the point cloud cluster. The spatial distribution characteristics are used to characterize the spatial sparsity of the point cloud cluster. The normal vector calculation unit 504 is used to determine the adaptive neighborhood value for each point cloud cluster based on the spatial distribution characteristics of the point cloud cluster, and to determine multiple local normal vectors of the point cloud cluster based on the adaptive neighborhood value. The local normal vector is used to characterize the normal vector of the plane formed by fitting the sampling point with the adaptive neighborhood value and the neighboring sampling points. The cluster identification unit 505 is used to determine the point cloud cluster corresponding to the illegal building based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster. The distribution characteristics are used to characterize the frequency of each cosine value interval in multiple cosine value intervals. The frequency is the number of direction cosines of local normal vectors contained in the cosine value interval. The graphics generation unit 506 is used to model and map textures of illegal buildings based on point cloud clusters and image data corresponding to illegal buildings in the target area, and generate graphics of abnormal land use areas.

[0116] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0117] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for generating anomalous land use area graphics based on LiDAR point clouds, characterized in that, The method includes: Acquire LiDAR point cloud data and image data of the target area, and preprocess the LiDAR point cloud data; The preprocessed LiDAR point cloud data is segmented to obtain multiple point cloud clusters. Based on the point cloud curvature and coordinates of each sampling point in the point cloud cluster, the spatial distribution characteristics of the point cloud cluster are determined. The spatial distribution characteristics are used to characterize the spatial sparsity of the point cloud cluster. For each point cloud cluster, an adaptive neighborhood value is determined based on the spatial distribution characteristics of the point cloud cluster, and multiple local normal vectors of the point cloud cluster are determined based on the adaptive neighborhood value. The local normal vectors are used to characterize the normal vectors of the plane formed by fitting the sampling point with the neighboring sampling points of the adaptive neighborhood value. Based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster, the point cloud cluster corresponding to the illegal building is determined. The distribution characteristics are used to characterize the frequency of each cosine value interval in multiple cosine value intervals. The frequency is the number of direction cosines of the local normal vectors contained in the cosine value interval. Based on the point cloud clusters corresponding to illegal buildings within the target area and the image data, model and map textures for illegal buildings to generate abnormal land use area graphics. Specifically, the spatial distribution characteristics of the point cloud cluster are determined based on the point cloud curvature and coordinates of each sampling point in the point cloud cluster, including: Based on the point cloud curvature of each sampling point in the point cloud cluster, determine the standard deviation of curvature and the average curvature. The curvature distribution characteristics of the point cloud cluster are determined based on the point cloud curvature of all sampling points in the point cloud cluster, the standard deviation of curvature, and the average curvature. Based on the curvature distribution characteristics and the coordinates of each sampling point in the point cloud cluster, the spatial distribution characteristics of the point cloud cluster are determined. Specifically, based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster, the point cloud clusters corresponding to illegal buildings are determined, including: Determine the direction cosine of the local normal vector corresponding to each sampling point in the point cloud cluster; the direction cosine is the X-axis direction cosine or the Y-axis direction cosine of the local normal vector; Determine the frequency of each interval of cosine values; For each cosine value interval, if the frequency is greater than a first threshold, the cosine value interval is determined to be a peak interval, and the first threshold is positively correlated with the number of sampling points included in the point cloud cluster; The number of peak intervals is determined, and if the number of peak intervals is less than a second threshold, the point cloud cluster is determined to be the point cloud cluster corresponding to the illegal building. The second threshold is determined based on the symmetry characteristics of the point cloud cluster.

2. The method for generating abnormal land use area graphics according to claim 1, characterized in that, Determining the adaptive neighborhood value based on the spatial distribution characteristics of the point cloud cluster includes: The spatial distribution characteristics of the point cloud cluster are normalized. The adaptive neighborhood value is determined based on the spatial distribution characteristics of the point cloud cluster after normalization, the preset maximum neighborhood value, and the preset minimum neighborhood value.

3. The method for generating abnormal land use area graphics according to claim 1, characterized in that, Based on the adaptive neighborhood values, multiple local normal vectors of the point cloud cluster are determined, including: For each sampling point in the point cloud cluster, a fitting plane between the sampling point and its neighboring sampling points of the adaptive neighborhood value is determined based on the least squares method; The normal vector of the fitted plane is determined as the local normal vector corresponding to the sampling point, thereby obtaining multiple local normal vectors of the point cloud cluster.

4. The method for generating abnormal land use area graphics according to claim 1, characterized in that, After identifying the point cloud cluster corresponding to the illegal building, the method further includes: Determine the extreme values ​​of the x-coordinate and y-coordinate of the sampling points in the point cloud cluster corresponding to the illegal building; The illegal building area is determined based on the extreme values ​​of the horizontal and vertical coordinates. The illegal building area is a spatial area formed by expanding a preset range based on the extreme values ​​of the horizontal and vertical coordinates. All sampling points located within the area of ​​the illegal building are identified as the point cloud cluster corresponding to the illegal building.

5. The method for generating abnormal land use area graphics according to claim 1, characterized in that, Based on the point cloud clusters corresponding to the illegal buildings within the target area and the image data, the illegal buildings are modeled and textures are mapped to generate the abnormal land use area graphics, including: The transformation matrix between the point cloud clusters corresponding to the illegal buildings and the image data is calculated by a feature matching algorithm, and the point cloud clusters corresponding to the illegal buildings are registered. The point cloud clusters corresponding to the registered illegal buildings are reconstructed to generate triangular mesh models; Based on the image data, texture mapping is performed on the triangular mesh model to obtain a textured illegal building model; The illegal building model is overlaid on the satellite map to generate the graphic of the abnormal land use area.

6. The method for generating abnormal land use area graphics according to claim 1, characterized in that, Preprocessing of the LiDAR point cloud data includes: The LiDAR point cloud data is processed using a preset filtering algorithm to identify and remove outliers, which are sampling points whose spatial distance from other sampling points exceeds a preset distance.

7. The method for generating abnormal land use area graphics according to claim 1, characterized in that, The preprocessed LiDAR point cloud data is segmented to obtain multiple point cloud clusters, including: Determine the point cloud curvature of each sampling point in the preprocessed LiDAR point cloud data; The preprocessed LiDAR point cloud data is divided into multiple point cloud clusters based on the point cloud curvature of the sampling points using a region growing segmentation algorithm.

8. A device for generating anomalous land use area graphics based on LiDAR point clouds, configured to implement the method of claim 1, characterized in that, include: The data acquisition unit is used to acquire LiDAR point cloud data and image data of the target area, and to preprocess the LiDAR point cloud data. The point cloud segmentation unit is used to segment the preprocessed LiDAR point cloud data to obtain multiple point cloud clusters. The feature analysis unit is used to determine the spatial distribution characteristics of the point cloud cluster, and the spatial distribution characteristics are used to characterize the spatial sparsity of the point cloud cluster. The normal vector calculation unit is used to determine an adaptive neighborhood value for each point cloud cluster based on the spatial distribution characteristics of the point cloud cluster, and to determine multiple local normal vectors of the point cloud cluster based on the adaptive neighborhood value. The local normal vectors are used to characterize the normal vector of the plane formed by fitting the sampling point with the neighboring sampling points of the adaptive neighborhood value. The cluster identification unit is used to determine the point cloud cluster corresponding to the illegal building based on the distribution characteristics of the direction cosines of multiple local normal vectors of each point cloud cluster. The distribution characteristics are used to characterize the frequency of each cosine value interval in multiple cosine value intervals. The frequency is the number of direction cosines of the local normal vectors contained in the cosine value interval. The graphics generation unit is used to model and map textures of illegal buildings based on the point cloud clusters corresponding to illegal buildings in the target area and the image data, and generate graphics of abnormal land use areas.

Images