Airborne laser radar point cloud individual tree segmentation method, device, equipment and medium

Abstract

The invention discloses an airborne laser radar point cloud individual tree segmentation method, device and equipment and a medium, and relates to the field of individual tree segmentation, and the method comprises the steps: carrying out the point cloud voxelization of a target point cloud, and obtaining a cuboid voxel; extracting die points in each cuboid voxel; generating a connected graph according to the mode points in each cuboid voxel and the corresponding neighbor mode point group; extracting an initial tree base point from the connected graph; determining a substance transportation path of a mode point in each cuboid voxel according to the connected graph and the initial tree base point; for any initial tree base point, determining the path number of the substance transportation paths passing through the initial tree base point; and screening the initial tree base points according to the path number to obtain a final tree base point, and determining a single tree segmentation result of the target forest according to the final tree base point and the material transportation path passing through the final tree base point, so that the precision and applicability of single tree segmentation can be improved.

Description

Technical Field

[0001] The present application relates to the field of single tree segmentation, and in particular to a method, device, equipment and medium for airborne laser radar point cloud single tree segmentation. Background Art

[0002] Forests are vital to maintaining ecological balance and sustainable development. Research on forests at the scale of individual trees helps reveal the complexity and diversity of ecosystems and provides important support for forest management and ecological research. Airborne LiDAR technology accurately depicts the three-dimensional structure of forests with high-resolution point clouds and is a common tool for obtaining information at the scale of individual trees in large areas.

[0003] Accurately segmenting individual trees in LiDAR point clouds is the key to quantifying tree structure. Related individual tree segmentation methods can be divided into two categories: top-down methods and bottom-up methods.

[0004] The top-down method analyzes trees based on canopy characteristics, including canopy height feature method and density feature method. The canopy height feature method identifies the tree tops through a local maximum filter and combines regional growing or watershed algorithms to segment the canopy. This type of method is suitable for coniferous forests with obvious canopy height characteristics. The canopy density feature method uses clustering algorithms to segment individual trees. Typical methods include Meanshift, K-means, and spectral clustering. This type of method is suitable for the segmentation of broad-leaved forests and understory vegetation, but is limited by the significance of point cloud distribution characteristics and has weak segmentation stability.

[0005] The bottom-up method detects individual trees by the position of the trunk. Trunk identification relies on point cloud density, strength, shape, and vertical morphological features. Compared with canopy features, trunk morphology has a higher similarity, and the bottom-up method is more robust to tree species and leaf conditions.

[0006] Although the bottom-up method is more versatile in different scenarios, it still faces the following challenges: the point cloud density method cannot eliminate the interference of understory vegetation; the intensity method is easily affected by the canopy occlusion on the point cloud intensity, and the performance is unstable; the shape method is only applicable to high-density, high-precision point clouds that can record the cross-sectional shape of the tree trunk; the vertical morphology method is not applicable enough in mixed forests. Therefore, how to improve the accuracy and applicability of single tree segmentation has become a problem that needs to be solved urgently. Summary of the invention

[0007] The purpose of this application is to provide a method, device, equipment and medium for airborne laser radar point cloud single tree segmentation to improve the accuracy and applicability of single tree segmentation.

[0008] To achieve the above objectives, this application provides the following solutions:

[0009] In a first aspect, the present application provides an airborne laser radar point cloud single tree segmentation method, comprising:

[0010] Obtaining a target point cloud; the target point cloud is obtained by scanning the target forest using an airborne laser radar;

[0011] voxelize the target point cloud to obtain cuboid voxels;

[0012] Extracting a mode point in each of the cuboid voxels; the mode point is a laser point closest to the center of gravity of the cuboid voxel;

[0013] Generate a connected graph according to the mode points in each of the cuboid voxels and the corresponding neighboring mode point groups; the neighboring mode point groups include: mode points in the neighboring cuboid voxels of the cuboid voxel where the mode point is located;

[0014] Extracting initial tree base points from the connected graph; the initial tree base points include: mode points in low-level voxels; the low-level voxels are cuboid voxels at a set level in the connected graph;

[0015] Determine the material transport path of each model point in each cuboid voxel according to the connectivity graph and the initial tree base point; the material transport path is the shortest path in the path set of the model point; the path set includes: the shortest path between the model point and each of the initial tree base points;

[0016] For any of the initial tree base points, determining the number of material transport paths passing through the initial tree base point;

[0017] The initial tree base points are screened according to the number of paths to obtain final tree base points, and the single tree segmentation result of the target forest is determined according to the final tree base points and the material transportation path passing through the final tree base points.

[0018] Optionally, obtain the target point cloud, specifically including:

[0019] An airborne LiDAR is used to collect the initial LiDAR point cloud of the target forest;

[0020] Extracting ground points from the initial point cloud of the laser radar using a cloth simulation filtering algorithm, and constructing a terrain model based on the ground points;

[0021] Calculating the height difference between the initial point cloud of the laser radar and the terrain model;

[0022] The height difference is used as the point cloud elevation, and the height information of the initial laser radar point cloud is normalized to obtain a height normalized point cloud, and the height normalized point cloud is used as the target point cloud.

[0023] Optionally, voxelizing the target point cloud to obtain cuboid voxels specifically includes:

[0024] The point cloud is voxelized according to the three-dimensional coordinates of each laser point in the target point cloud to obtain a cuboid voxel; the expression of the cuboid voxel is:

[0025]

[0026] Where (r,c,l) represents the row, column and layer number of the cuboid voxel to which the laser point belongs, r represents the row number, c represents the column number, and l represents the layer number; int represents the rounding function; (x,y,z) represents the three-dimensional coordinates of the laser point, x represents the x-axis coordinate value, y represents the y-axis coordinate value, and z represents the z-axis coordinate value; x represents the x-axis coordinate value, y represents the y-axis coordinate value, and z represents the z-axis coordinate value; min Indicates the minimum x-axis coordinate value of the target point cloud, y min Indicates the minimum y-axis coordinate value of the target point cloud, z min Indicates the minimum z-axis coordinate value of the target point cloud; v xy Indicates the horizontal size of the cuboid voxel; v z Indicates the vertical size of the cuboid voxel.

[0027] Optionally, extracting the mode points in each of the cuboid voxels specifically includes:

[0028] For any of the cuboid voxels, the Euclidean distances between each laser point in the cuboid voxel and the center of gravity of the cuboid voxel are calculated, and the laser point with the smallest Euclidean distance is determined as the mode point in the cuboid voxel.

[0029] Optionally, generating a connectivity graph according to the mode points and the corresponding neighboring mode point groups in each of the cuboid voxels specifically includes:

[0030] For any mode point in the cuboid voxel, determine the mode points in 26 neighboring cuboid voxels of the cuboid voxel where the mode point is located as a neighboring mode point group of the mode point;

[0031] For any mode point in the cuboid voxel, the mode point and each mode point in the corresponding neighboring mode point group are connected to each other to generate a topological map of the mode point;

[0032] The connectivity graph is determined based on the topological graph of the mode points within all cuboid voxels.

[0033] Optionally, determining the material transport path of the model point in each of the cuboid voxels according to the connectivity graph and the initial tree base point specifically includes:

[0034] For any mode point in the cuboid voxel, the shortest path between the mode point and each of the initial tree base points is calculated using the Dijkstra algorithm according to the connectivity graph to obtain a path set of the mode point;

[0035] For any model point within the cuboid voxel, the shortest path is selected from the path set to obtain the material transport path of the model point.

[0036] Optionally, the initial tree base points are screened according to the number of paths to obtain final tree base points, and a single tree segmentation result of the target forest is determined according to the final tree base points and a material transportation path passing through the final tree base points, specifically including:

[0037] Determine the initial tree base point whose number of paths exceeds a set threshold as the final tree base point;

[0038] For any of the final tree base points, all material transport paths passing through the final tree base point are determined as a target path set for the final tree base point;

[0039] For any of the final tree base points, the cuboid voxels to which the model points corresponding to all the material transport paths in the target path set belong are determined as the target voxels corresponding to the final tree base point;

[0040] For any of the final tree base points, a single tree segmentation result of the final tree base point is generated according to the point cloud in the target voxel; the single tree segmentation results of all the final tree base points are used as the single tree segmentation results of the target forest.

[0041] In a second aspect, the present application provides an airborne laser radar point cloud single tree segmentation device, comprising:

[0042] A point cloud acquisition module is used to acquire a target point cloud; the target point cloud is obtained by scanning the target forest using an airborne laser radar;

[0043] A point cloud voxelization module, used for performing point cloud voxelization on the target point cloud to obtain cuboid voxels;

[0044] A mode point extraction module, used to extract a mode point in each of the cuboid voxels; the mode point is a laser point closest to the center of gravity of the cuboid voxel;

[0045] A connectivity graph generation module, used to generate a connectivity graph according to the mode points in each of the cuboid voxels and the corresponding neighboring mode point groups; the neighboring mode point groups include: mode points in the neighboring cuboid voxels of the cuboid voxel where the mode point is located;

[0046] An initial tree base point extraction module is used to extract initial tree base points from the connected graph; the initial tree base points include: mode points in low-level voxels; the low-level voxels are cuboid voxels at a set layer number in the connected graph;

[0047] A material transport path determination module is used to determine the material transport path of each model point in each of the cuboid voxels according to the connectivity graph and the initial tree base points; the material transport path is the shortest path in the path set of the model point; the path set includes: the shortest path between the model point and each of the initial tree base points;

[0048] A path quantity determination module, used for determining the path quantity of the material transportation path passing through any of the initial tree base points;

[0049] The single tree segmentation result determination module is used to screen the initial tree base points according to the number of paths to obtain the final tree base points, and determine the single tree segmentation result of the target forest according to the final tree base points and the material transportation path passing through the final tree base points.

[0050] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described airborne laser radar point cloud single tree segmentation methods.

[0051] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described airborne laser radar point cloud single tree segmentation methods.

[0052] According to the specific embodiments provided in this application, this application has the following technical effects:

[0053] The present application provides an airborne laser radar point cloud single tree segmentation method, device, equipment and medium. The method voxelizes the laser radar point cloud to obtain rectangular voxels, extracts the laser point closest to the center of gravity of each rectangular voxel as the model point, and generates a connectivity graph based on the model points in each rectangular voxel and the corresponding neighboring model point groups. The constructed connectivity graph has stronger vertical connectivity and is used for single tree segmentation, which can significantly improve the accuracy of single tree segmentation. The method extracts the initial tree base point from the connectivity graph, determines the material transportation path of the model point in each rectangular voxel according to the connectivity graph and the initial tree base point, and performs single tree segmentation based on the material transportation path. The method is applicable to a variety of complex forest scenes and ensures the high applicability of single tree segmentation under different conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 This is an application environment diagram of an airborne laser radar point cloud single tree segmentation method in one embodiment of the present application;

[0056] Figure 2 A schematic diagram of a flow chart of a method for segmenting a single tree in an airborne laser radar point cloud provided in one embodiment of the present application;

[0057] Figure 3 A key step effect diagram provided for an embodiment of the present application;

[0058] Figure 4 A schematic diagram of single tree segmentation results for different forest scenes provided in an embodiment of the present application;

[0059] Figure 5 A schematic diagram of functional modules of an airborne laser radar point cloud single tree segmentation device provided in another embodiment of the present application;

[0060] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0062] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0063] The airborne laser radar point cloud single tree segmentation method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the target point cloud to the server 104. After the server 104 receives the target point cloud, for the target point cloud, the server 104 voxelizes the target point cloud to obtain rectangular voxels; extracts the model points in each rectangular voxel; generates a connected graph based on the model points in each rectangular voxel and the corresponding neighboring model point groups; extracts the initial tree base point from the connected graph; determines the material transportation path of the model point in each rectangular voxel according to the connected graph and the initial tree base point; for any initial tree base point, determines the number of paths of the material transportation path passing through the initial tree base point; screens the initial tree base point according to the number of paths to obtain the final tree base point, and determines the single tree segmentation result of the target forest according to the final tree base point and the material transportation path passing through the final tree base point.

[0064] The server 104 may feed back the obtained single tree segmentation result of the target forest to the terminal 102. In addition, in some embodiments, the airborne laser radar point cloud single tree segmentation method may also be implemented by the server 104 or the terminal 102 alone, such as the terminal 102 may directly process the target point cloud, or the server 104 may obtain the target point cloud from the data storage system and process the target point cloud.

[0065] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0066] In an exemplary embodiment, Figure 2 As shown, a method for segmenting a single tree in an airborne laser radar point cloud is provided. The method is executed by a computer device, and specifically can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps 201 to 208. Among them:

[0067] Step 201, obtaining a target point cloud.

[0068] The target point cloud is obtained by scanning the target forest using an airborne laser radar, and the target point cloud includes multiple laser points.

[0069] Step 202 , voxelize the target point cloud to obtain cuboid voxels.

[0070] Step 203: extracting the mode points in each of the cuboid voxels.

[0071] The mode point is a laser point closest to the center of gravity of the cuboid voxel.

[0072] Step 204: Generate a connectivity graph based on the mode points in each of the cuboid voxels and the corresponding neighboring mode point groups.

[0073] The neighboring mode point group includes mode points in neighboring rectangular parallelepiped voxels of the rectangular parallelepiped voxel where the mode point is located.

[0074] Step 205: extracting initial tree base points from the connected graph.

[0075] The initial tree base points include: mode points in low-level voxels; the low-level voxels are cuboid voxels at a set layer number in the connected graph.

[0076] Step 206, determining the material transport path of the model point in each of the cuboid voxels according to the connectivity graph and the initial tree base points.

[0077] The material transport path is the shortest path in the path set of the model point; the path set includes: the shortest path between the model point and each of the initial tree base points.

[0078] Step 207: for any of the initial tree base points, determine the number of material transport paths that pass through the initial tree base point.

[0079] The purpose of determining the number of paths is to evaluate the cumulative strength of each initial tree base point to facilitate the screening of subsequent tree base points.

[0080] Step 208, screening the initial tree base points according to the number of paths to obtain final tree base points, and determining the single tree segmentation result of the target forest according to the final tree base points and the material transportation paths passing through the final tree base points.

[0081] Implementing the above steps 201 to 208 can improve the accuracy and applicability of single tree segmentation.

[0082] In another exemplary embodiment of the present application, step 201 specifically includes:

[0083] An airborne laser radar is used to collect the laser radar initial point cloud of the target forest; a cloth simulation filtering algorithm is used to extract ground points from the laser radar initial point cloud, and a terrain model is constructed based on the ground points; the height difference between the laser radar initial point cloud and the terrain model is calculated; the height difference is used as the point cloud elevation, and the height information of the laser radar initial point cloud is normalized to obtain a height normalized point cloud, and the height normalized point cloud is used as the target point cloud. The normalization process is: the height difference is used to replace the height information of the laser radar initial point cloud.

[0084] In another exemplary embodiment of the present application, step 202 specifically includes:

[0085] The point cloud is voxelized according to the three-dimensional coordinates of each laser point in the target point cloud to obtain a cuboid voxel, which forms the basic unit for model point extraction. The expression of the cuboid voxel is:

[0086]

[0087] Where (r,c,l) represents the row, column and layer number of the cuboid voxel to which the laser point belongs, r represents the row number, c represents the column number, and l represents the layer number; int represents the rounding function; (x,y,z) represents the three-dimensional coordinates of the laser point, x represents the x-axis coordinate value, y represents the y-axis coordinate value, and z represents the z-axis coordinate value; x represents the x-axis coordinate value, y represents the y-axis coordinate value, and z represents the z-axis coordinate value; min ,y min and z min Indicates the minimum x, y, and z values ​​of the target point cloud, that is, x min Indicates the minimum x-axis coordinate value of the target point cloud, y min Indicates the minimum y-axis coordinate value of the target point cloud, z min Indicates the minimum z-axis coordinate value of the target point cloud; v xy Indicates the horizontal size of the cuboid voxel; v z Indicates the vertical size of the cuboid voxel.

[0088] In another exemplary embodiment of the present application, step 203 specifically includes:

[0089] For any of the cuboid voxels, the Euclidean distances between each laser point in the cuboid voxel and the center of gravity of the cuboid voxel are calculated, and the laser point with the smallest Euclidean distance is determined as the mode point in the cuboid voxel, thereby providing a basis for constructing a connectivity graph.

[0090] In this embodiment, in each cuboid voxel, the laser point closest to the center of gravity is selected as the mode point, and its calculation formula is as follows:

[0091] argmin||p i -c|| (2)

[0092] Among them, argmin represents the index function of the voxel point v, ||·|| represents the Euclidean distance, p i is the coordinate of the i-th laser point, and c represents the coordinate of the centroid point. All voxel points are stored in the voxel point set V, that is, v ∈ V.

[0093] In another exemplary embodiment of the present application, step 204 specifically includes:

[0094] For any voxel point within any of the cuboid voxels, the voxel points within the 26-nearest-neighbor cuboid voxels of the cuboid voxel where the voxel point is located are determined as the neighboring voxel point group of the voxel point; for any voxel point within any of the cuboid voxels, the voxel point and each voxel point in the corresponding neighboring voxel point group are connected to each other to generate a topological graph of the voxel point; a connected graph is determined according to the topological graphs of all voxel points within all cuboid voxels. The connected graph is used to describe the topological relationship between point clouds.

[0095] In this embodiment, the calculation formula for any neighboring voxel point v' in the neighboring voxel point group is as follows:

[0096] {(r', c', l')||r - r'| ≤ 1, |c - c'| ≤ 1, |l - l'| ≤ 1, (r', c', l') ≠ (r, c, l)} (3)

[0097] Among them, (r, c, l) and (r', c', l') respectively represent the row, column, and layer numbers of v and v'. The weight of the connecting edge e between v and v' is the Euclidean distance ||v - v'|| between the two points. All connecting edges are stored in the edge set E, that is, e ∈ E.

[0098] In another exemplary embodiment of the present application, in step 205, the voxel points of the lower-layer voxels are extracted from the connected graph as the initial tree base points to calibrate the potential trunk starting points. The lower-layer voxels are the cuboid voxels in the second layer from bottom to top in the connected graph.

[0099] In another exemplary embodiment of the present application, step 206 specifically includes:

[0100] (1) For any voxel point within any of the cuboid voxels, according to the connected graph, the Dijkstra algorithm is used to calculate the shortest path between the voxel point and each of the initial tree base points, and a path set of the voxel point is obtained.

[0101] The shortest path between the voxel point and the initial tree base point is determined based on the path length d(v') from the initial tree base point to the neighboring voxel point v'. The calculation formula for d(v') is as follows:

[0102] d(v') = min(d(v) + ||v - v'||) (4)

[0103] v' is the neighboring module point of v, d(v) and d(v') represent the path lengths from the initial tree base point to the module points v and v', and ||v-v'|| represents the Euclidean distance between the module points v and v'. In order to improve efficiency, this embodiment only calculates the shortest path between each module point and the initial tree base point within a horizontal distance of 10m.

[0104] (2) For any model point in the cuboid voxel, the shortest path is selected from the path set to obtain the material transport path of the model point, that is, the potential path of the model point. The calculation formula of the corresponding initial tree base point is:

[0105] argmin d(s i ) (5)

[0106] Among them, argmin is the index of the initial tree base point corresponding to a modulus point, d(s i ) is the modulus point to the initial tree base point s i The length of the path between .

[0107] In another exemplary embodiment of the present application, step 208 specifically includes:

[0108] (1) The initial tree base point whose number of paths exceeds a set threshold is determined as the final tree base point.

[0109] (2) For any of the final tree base points, all material transportation paths passing through the final tree base point are determined as the target path set of the final tree base point.

[0110] (3) For any of the final tree base points, the cuboid voxels to which the model points corresponding to all the material transport paths in the target path set belong are determined as the target voxels corresponding to the final tree base point.

[0111] (4) For any of the final tree base points, a single tree segmentation result of the final tree base point is generated according to the point cloud in the target voxel; the single tree segmentation results of all the final tree base points are used as the single tree segmentation results of the target forest.

[0112] The airborne laser radar point cloud single tree segmentation method of the above-mentioned embodiment, first, organizes the point cloud through rectangular voxels, retains the near-center-of-gravity point of each voxel as the composition vertex, and generates connecting edges based on the neighboring voxel vertices, which simplifies the number of composition vertices and improves the composition efficiency. Secondly, the constructed connectivity graph has stronger vertical connectivity, which significantly improves the accuracy of single tree segmentation. In addition, the material transport path theory of the algorithm has little effect on tree species, leaf conditions, understory vegetation and tree density, and is suitable for a variety of complex forest scenes, ensuring high applicability and accuracy of single tree segmentation under different conditions. Finally, the cumulative intensity characteristics of the tree base points are related to the crown volume, which can effectively alleviate the interference of understory vegetation and further improve the segmentation accuracy.

[0113] The following describes an implementation of a single tree segmentation method for airborne lidar point cloud and illustrates the effectiveness of the method.

[0114] The implementation of the airborne laser radar point cloud single tree segmentation method includes:

[0115] Step 1: Normalize the height information of the airborne lidar point cloud.

[0116] Step 2: Use the normalized point cloud to divide it into cuboid voxels to form the basic unit for model point extraction. The calculation formula of the cuboid voxel is shown in formula (1), which will not be repeated here.

[0117] Step 3: In each voxel, extract the laser point closest to the center of gravity as the model point to provide a basis for the construction of the connectivity graph. The calculation formula of the model point is shown in formula (2), which will not be repeated here.

[0118] Step 4: Based on the position of the model point, connect the model points in its 26 nearest neighbor voxels to generate a connected graph to describe the topological relationship between the points. The 26 nearest neighbor model points v' of each model point v are identified based on the row and column layer numbers of the voxels. The calculation formula is shown in formula (3), which will not be repeated here.

[0119] Step 5: Extract the model points of low-level voxels from the connectivity graph as the initial tree base points to calibrate potential trunk starting points.

[0120] Step 6: Using the initial tree base point and the connectivity graph, calculate the shortest path from the module point to the initial tree base point to form a path set of the module point. The formula for calculating the shortest path from each module point to the initial tree base point using the Dijkstra algorithm is shown in formula (4), which will not be repeated here.

[0121] Step 7: Select the shortest path in the path set and define it as the material transportation path between the model point and the initial tree base point. After obtaining the potential path of each model point, the shortest path in the path set is retained as the material transportation path. The corresponding initial tree base point calculation formula is shown in formula (5), which will not be repeated here.

[0122] Step 8: According to the material transport path, count the number of paths for each initial tree base point and evaluate the cumulative strength of the initial tree base point.

[0123] Step 9: Filter the initial tree base points and their paths whose path number exceeds the set threshold, remove redundant paths and initial tree base points, and obtain the final tree base points.

[0124] Step 10: Based on the screening results, extract the laser points in the voxels corresponding to the path to generate the final single tree segmentation results. The single tree segmentation results can be output as a point cloud file (*.las) for more convenient use.

[0125] The above-mentioned airborne laser radar point cloud single tree segmentation method based on material transportation path realizes accurate segmentation of single trees by identifying the material transportation path in the point cloud. In the forest point cloud, the starting point and end point of the material transportation path correspond to the root point and crown of the tree respectively. Using this path characteristic, the root point of the tree and its corresponding crown range are reversely deduced to complete the single tree segmentation. Since the material transportation path theory has a low dependence on tree species, leaf conditions, understory vegetation and tree density, this embodiment has good versatility and is applicable to a variety of complex forest scenes.

[0126] The effect diagram obtained based on the above steps is as follows Figure 3 As shown. Among them, Figure 3 Part (a) shows the lidar point cloud; Figure 3 Construction of the partially connected graph in (b), where the black dots and gray lines represent the model points and edges of the connected graph, respectively; Figure 3 Part (c) shows the identification of material transport paths. The black rectangular box represents the tree base, and the edge represents the material transport path from the model point to the tree base. The edges pointing to the same tree base are coded with a unified color. Figure 3 Part (d) shows the single tree segmentation result. The single tree point cloud is rendered in different colors, and the gray points represent the understory vegetation.

[0127] Figure 4 The results of the method for segmenting single trees using airborne laser radar point cloud based on material transport paths are shown in this embodiment. Compared with existing methods, such as the PCS (Point Cloud Segmentation) algorithm and the AMS3D algorithm, the method of this embodiment shows higher versatility in different scenarios and significantly improves the accuracy of single tree segmentation. The method of this embodiment can achieve high accuracy in forest scenarios with different tree species, leaf conditions, understory vegetation, tree density, and point density, and has strong practical value.

[0128] Based on the same inventive concept, the embodiment of the present application also provides an airborne laser radar point cloud single tree segmentation device for implementing the airborne laser radar point cloud single tree segmentation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more airborne laser radar point cloud single tree segmentation device embodiments provided below can refer to the limitations of the airborne laser radar point cloud single tree segmentation method above, and will not be repeated here.

[0129] In an exemplary embodiment, Figure 5 As shown, an airborne laser radar point cloud single tree segmentation device is provided, comprising:

[0130] The point cloud acquisition module 501 is used to acquire a target point cloud; the target point cloud is obtained by scanning the target forest using an airborne laser radar.

[0131] The point cloud voxelization module 502 is used to perform point cloud voxelization on the target point cloud to obtain cuboid voxels.

[0132] The mode point extraction module 503 is used to extract the mode point in each of the cuboid voxels; the mode point is the laser point closest to the center of gravity of the cuboid voxel.

[0133] The connectivity graph generation module 504 is used to generate a connectivity graph according to the mode points in each of the cuboid voxels and the corresponding neighboring mode point groups; the neighboring mode point groups include: mode points in the neighboring cuboid voxels of the cuboid voxel where the mode point is located.

[0134] The initial tree base point extraction module 505 is used to extract the initial tree base points from the connected graph; the initial tree base points include: model points in low-level voxels; the low-level voxels are cuboid voxels at a set layer number in the connected graph.

[0135] The material transport path determination module 506 is used to determine the material transport path of each model point in the rectangular voxel according to the connectivity graph and the initial tree base point; the material transport path is the shortest path in the path set of the model point; the path set includes: the shortest path between the model point and each of the initial tree base points.

[0136] The path quantity determination module 507 is used to determine the path quantity of the material transportation path passing through any of the initial tree base points.

[0137] The single tree segmentation result determination module 508 is used to screen the initial tree base points according to the number of paths to obtain final tree base points, and determine the single tree segmentation result of the target forest according to the final tree base points and the material transportation path passing through the final tree base points.

[0138] This embodiment is applicable to multiple types of forest scenes with different tree species, leaf states, understory vegetation, tree density and point density. It can significantly improve the accuracy of forest structure depiction at the scale of a single tree, help to deeply analyze the complexity and diversity of forest ecosystems, and provide important support for forest management and ecological research.

[0139] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store target point clouds. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for segmenting single trees of airborne laser radar point clouds is implemented.

[0140] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0141] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0142] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0143] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0144] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0145] The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, etc., without limitation.

[0146] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0147] In this article, specific examples are used to elaborate on the principles and implementation manners of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A method for segmenting single trees from airborne laser radar point clouds, characterized in that: The airborne laser radar point cloud single tree segmentation method comprises: Obtaining a target point cloud; the target point cloud is obtained by scanning the target forest using an airborne laser radar; voxelize the target point cloud to obtain cuboid voxels; Extracting a mode point in each of the cuboid voxels; the mode point is a laser point closest to the center of gravity of the cuboid voxel; Generate a connected graph according to the mode points in each of the cuboid voxels and the corresponding neighboring mode point groups; the neighboring mode point groups include: mode points in the neighboring cuboid voxels of the cuboid voxel where the mode point is located; Extracting initial tree base points from the connected graph; the initial tree base points include: mode points in low-level voxels; the low-level voxels are cuboid voxels at a set level in the connected graph; Determine the material transport path of each model point in each cuboid voxel according to the connectivity graph and the initial tree base point; the material transport path is the shortest path in the path set of the model point; the path set includes: the shortest path between the model point and each of the initial tree base points; For any of the initial tree base points, determining the number of material transport paths passing through the initial tree base point; The initial tree base points are screened according to the number of paths to obtain final tree base points, and the single tree segmentation result of the target forest is determined according to the final tree base points and the material transportation path passing through the final tree base points.

2. The method for segmenting single trees from airborne laser radar point cloud according to claim 1, characterized in that: Get the target point cloud, including: An airborne LiDAR is used to collect the initial LiDAR point cloud of the target forest; Extracting ground points from the initial point cloud of the laser radar using a cloth simulation filtering algorithm, and constructing a terrain model based on the ground points; Calculating the height difference between the initial point cloud of the laser radar and the terrain model; The height difference is used as the point cloud elevation, and the height information of the initial laser radar point cloud is normalized to obtain a height normalized point cloud, and the height normalized point cloud is used as the target point cloud.

3. The method for segmenting single trees from airborne laser radar point cloud according to claim 1, characterized in that: The target point cloud is voxelized to obtain a cuboid voxel, which specifically includes: The point cloud is voxelized according to the three-dimensional coordinates of each laser point in the target point cloud to obtain a cuboid voxel; the expression of the cuboid voxel is: Where (r,c,l) represents the row, column and layer number of the cuboid voxel to which the laser point belongs, r represents the row number, c represents the column number, and l represents the layer number; int represents the rounding function; (x,y,z) represents the three-dimensional coordinates of the laser point, x represents the x-axis coordinate value, y represents the y-axis coordinate value, and z represents the z-axis coordinate value; x represents the x-axis coordinate value, y represents the y-axis coordinate value, and z represents the z-axis coordinate value; min Indicates the minimum x-axis coordinate value of the target point cloud, y min Indicates the minimum y-axis coordinate value of the target point cloud, z min Indicates the minimum z-axis coordinate value of the target point cloud; v xy Indicates the horizontal size of the cuboid voxel; v z Indicates the vertical size of the cuboid voxel.

4. The method for segmenting single trees from airborne laser radar point cloud according to claim 1, characterized in that: Extracting the mode points in each of the cuboid voxels specifically includes: For any of the cuboid voxels, the Euclidean distances between each laser point in the cuboid voxel and the center of gravity of the cuboid voxel are calculated, and the laser point with the smallest Euclidean distance is determined as the mode point in the cuboid voxel.

5. The method for segmenting single trees from airborne laser radar point cloud according to claim 1, characterized in that: Generate a connectivity graph according to the mode points and the corresponding neighboring mode point groups in each of the cuboid voxels, specifically including: For any mode point in the cuboid voxel, determine the mode points in 26 neighboring cuboid voxels of the cuboid voxel where the mode point is located as a neighboring mode point group of the mode point; For any mode point in the cuboid voxel, the mode point and each mode point in the corresponding neighboring mode point group are connected to each other to generate a topological map of the mode point; The connectivity graph is determined based on the topological graph of the mode points within all cuboid voxels.

6. The method for segmenting single trees from airborne laser radar point cloud according to claim 1, characterized in that: Determining the material transport path of the model point in each of the cuboid voxels according to the connectivity graph and the initial tree base point specifically includes: For any mode point in the cuboid voxel, the shortest path between the mode point and each of the initial tree base points is calculated using the Dijkstra algorithm according to the connectivity graph to obtain a path set of the mode point; For any model point in the cuboid voxel, the shortest path is selected from the path set to obtain the material transport path of the model point.

7. The method for segmenting single trees from airborne laser radar point cloud according to claim 1, characterized in that: The initial tree base points are screened according to the number of paths to obtain final tree base points, and the single tree segmentation result of the target forest is determined according to the final tree base points and the material transportation path passing through the final tree base points, specifically including: Determine the initial tree base point whose number of paths exceeds a set threshold as the final tree base point; For any of the final tree base points, all material transport paths passing through the final tree base point are determined as a target path set for the final tree base point; For any of the final tree base points, the cuboid voxels to which the model points corresponding to all the material transport paths in the target path set belong are determined as the target voxels corresponding to the final tree base point; For any of the final tree base points, a single tree segmentation result of the final tree base point is generated according to the point cloud in the target voxel; the single tree segmentation results of all the final tree base points are used as the single tree segmentation results of the target forest.

8. An airborne laser radar point cloud single tree segmentation device, characterized in that: The airborne laser radar point cloud single tree segmentation device comprises: A point cloud acquisition module is used to acquire a target point cloud; the target point cloud is obtained by scanning the target forest using an airborne laser radar; A point cloud voxelization module, used for performing point cloud voxelization on the target point cloud to obtain cuboid voxels; A mode point extraction module, used to extract a mode point in each of the cuboid voxels; the mode point is a laser point closest to the center of gravity of the cuboid voxel; A connectivity graph generation module, used to generate a connectivity graph according to the mode points in each of the cuboid voxels and the corresponding neighboring mode point groups; the neighboring mode point groups include: mode points in the neighboring cuboid voxels of the cuboid voxel where the mode point is located; An initial tree base point extraction module is used to extract initial tree base points from the connected graph; the initial tree base points include: mode points in low-level voxels; the low-level voxels are cuboid voxels at a set layer number in the connected graph; A material transport path determination module is used to determine the material transport path of each model point in each of the cuboid voxels according to the connectivity graph and the initial tree base points; the material transport path is the shortest path in the path set of the model point; the path set includes: the shortest path between the model point and each of the initial tree base points; A path quantity determination module, used for determining the path quantity of the material transportation path passing through any of the initial tree base points; The single tree segmentation result determination module is used to screen the initial tree base points according to the number of paths to obtain the final tree base points, and determine the single tree segmentation result of the target forest according to the final tree base points and the material transportation path passing through the final tree base points.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the airborne laser radar point cloud single tree segmentation method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the airborne laser radar point cloud single tree segmentation method described in any one of claims 1 to 7 is implemented.

Images