Root system point cloud skeleton extraction method, system and device

Abstract

The invention discloses a root system point cloud skeleton extraction method, system and device, and the method comprises the steps: obtaining a plant root system point cloud, carrying out the point cloud division of the plant root system point cloud, and forming a root system voxel grid set; a root system voxel graph is constructed, connection is performed based on the spatial position of the current root system voxel grid and the spatial position of the adjacent root system voxel grid to obtain the edge of the root system voxel graph, and the weight of the root system voxel graph is obtained by constructing a point cloud search space and obtaining the point cloud density in the point cloud search space; based on the root system voxel graph, constructing a root system voxel spanning tree, judging noise in the root system voxel graph according to the root system voxel spanning tree, and filtering to obtain an initial root system skeleton graph; acquiring root system connection data of the current root system voxel grid, and filtering the root system voxel grid corresponding to the burrs through judgment to obtain a root system skeleton diagram. According to the method, skeleton extraction is performed from the point cloud data, skeleton misjudgment is reduced, and the problems of smear, noisy points, burrs and the like on the surface of a root system are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud extraction, and particularly to a method, system and device for extracting the root system point cloud skeleton. Background Art

[0002] Point cloud skeleton extraction is a method for extracting the skeleton structure from point cloud data, which can be used in fields such as object shape analysis and 3D model simplification. Existing point cloud skeleton extraction methods include the skeleton extraction method based on Laplace Contraction, the skeleton extraction method based on Rotational Symmetry Axis, and the skeleton extraction method based on L1-Medial. Among them, the skeleton extraction method based on Laplace Contraction uses the Laplace operator to process the point cloud data to obtain an estimate. This method can be used to process complex geometric shapes, but is susceptible to point cloud noise; the skeleton extraction method based on Rotational Symmetry Axis extracts the rotational symmetry axis of the object in a certain direction and takes the symmetry axis as part of the skeleton. This method has a good effect on objects with obvious rotational symmetry, but the calculation process of the symmetry axis in this method is relatively complex; the skeleton extraction method based on L1-Medial extracts the geometric features of the point cloud data using the idea of spatial median as the skeleton. This method can process point cloud data with complex geometric shapes, but requires adjusting many parameters during the extraction process.

[0003] Due to reasons such as the uneven distribution of the root system point cloud, the presence of trailing shadows on the root surface, the surface not being smooth, and the presence of discrete noise points near the root system, there are many misjudgments in the root system skeletons obtained by existing root system extraction methods, and there are many parameters that need to be finely adjusted in the existing methods, resulting in a low accuracy rate for root system skeleton extraction. Summary of the Invention

[0004] The present invention aims at the deficiencies in the prior art and provides a method, system and device for extracting the root system point cloud skeleton.

[0005] To solve the above technical problems, the present invention is solved by the following technical solutions: A method for extracting the root system point cloud skeleton includes the following steps: Obtain the plant root system point cloud and perform point cloud partitioning to form a root system voxel grid set; Construct a root system voxel graph, wherein, based on the spatial positions of the current root system voxel grid and the adjacent root system voxel grids, the edges of the root system voxel graph are connected, and the weights of the root system voxel graph are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; Based on the weights of the root voxel map and the root voxel map, construct a root voxel spanning tree using the minimum spanning tree algorithm, and then determine whether there is noise in the root voxel map, and filter the edges corresponding to the noise to obtain an initial root skeleton map; Obtain the connection relationship between the current root voxel grid and adjacent root voxel grids in the initial root skeleton map to obtain root connection data, and determine whether there are burrs in the initial root skeleton map, and filter the root voxel grids corresponding to the burrs to obtain a root skeleton map.

[0006] As an implementable manner, the point cloud division of the plant root system point cloud includes the following steps: Obtain the plant root system point cloud and determine the three-dimensional space boundary of the plant root system point cloud, where the three-dimensional space boundary includes a horizontal boundary, a vertical boundary, and a vertical boundary; Divide the plant root system point cloud within the three-dimensional space boundary according to a preset grid threshold to obtain an initial root voxel grid set; Obtain the number of point clouds included in all the initial root voxel grids in the initial root voxel grid set to obtain a point cloud quantity set, set a point cloud quantity threshold, obtain the number of point clouds in the point cloud quantity set that do not meet the point cloud quantity threshold, and filter the corresponding initial root voxel grids to obtain a filtered voxel grid set; In the filtered voxel grid set, if the filtered voxel grid set still has connectivity after deleting the current filtered voxel grid, then delete the current filtered voxel grid, and obtain a root voxel grid set through iterative deletion.

[0007] As an implementable manner, the step of connecting based on the spatial position of the current root voxel grid and the spatial position of the adjacent root voxel grid to obtain the edges of the root voxel map includes the following steps: Obtain the spatial position of the root voxel grid through the center position of the root voxel grid; Preset a search threshold, construct a search space for the current root voxel grid based on the search threshold, and determine whether there are adjacent root voxel grids in the search space; If there are, connect the spatial position of the current root voxel grid and the spatial position of the adjacent root voxel grid to obtain the edges of the root voxel map.

[0008] As an implementable manner, the step of obtaining the weights of the root voxel map by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space includes the following steps: Take the spatial position of the root voxel grid as the center of the sphere and the preset grid threshold as the diameter to construct a point cloud search area; Obtain the number of plant root system point clouds within the point cloud search area and perform density calculation to obtain the root point cloud density; Normalize the root system point cloud density to obtain the normalized point cloud density; Calculate based on adjacent normalized point cloud densities to obtain the weights of the root system voxel map; Among them, the normalized point cloud density and the weights of the root system voxel map are expressed as follows:

[0009]

[0010] Among them, , represents the maximum value of the root system point cloud density, represents the root system point cloud density, represents the number of root system voxel grids, represents the th normalized point cloud density, represents the th root system voxel grid, represents the th root system voxel grid and the th root system voxel grid, the weight of the root system voxel map between them, represents the th normalized point cloud density.

[0011] As an implementable manner, based on the weights of the root system voxel map and the root system voxel map, construct a root system voxel spanning tree based on the minimum spanning tree algorithm, and then determine whether there is noise in the root system voxel map, and filter the edges corresponding to the noise to obtain an initial root system skeleton map, including the following steps: Sort the root system voxel map according to the weights of the root system voxel map to obtain a sorted voxel map; Initialize the root system generation set, and determine whether the root system voxel grid corresponding to the edge of the current sorted voxel map belongs to the root system generation set. If not, add the edge of the current sorted voxel map to the root system generation set; If it belongs, filter the edge of the current sorted voxel map, traverse the sorted voxel map until the edges of the root system generation set can connect all the root system voxel grids in the sorted voxel map, and the sum of the weights of the root system voxel generation set is the smallest, to obtain the root system voxel spanning tree; Filter the edges in the root system voxel map that do not belong to the root system voxel spanning tree to obtain the initial root system skeleton map.

[0012] As an implementable manner, the steps for obtaining the connection relationship between the current root system voxel grid and adjacent root system voxel grids to obtain root system connection data include the following steps: Initialize the root system connection data of the current root system voxel grid in the initial root system skeleton map to 0; Traverse the edges of the initial root system skeleton graph, obtain the number of edges of the initial root system skeleton graph connected to the current root system voxel grid, and obtain the root system connection data corresponding to all root system voxel grids.

[0013] As an implementable manner, determining whether there are burrs in the initial root system skeleton graph and filtering the root system voxel grids corresponding to the burrs to obtain the root system skeleton graph includes the following steps: Obtain the root system connection data of the current root system voxel grid in the initial root system skeleton graph to obtain the first voxel data, and obtain the root system connection data of the adjacent root system voxel grid to obtain the second voxel data; Based on the first voxel data and the second voxel data, determine whether the root system voxel grid meets the burr condition. If it meets, it is determined as a burr. Among them, the burr condition is at least: the first voxel data is 1 and the second voxel data is greater than 2; Traverse the initial root system skeleton graph, filter the root system voxel grids corresponding to the burrs, and obtain the root system skeleton graph. A root system point cloud skeleton extraction system includes a point cloud division module, a voxel graph construction module, a noise filtering module, and a burr filtering module; The point cloud division module obtains the plant root system point cloud and performs point cloud division to form a set of root system voxel grids; The voxel graph construction module constructs a root system voxel graph. Among them, the edges of the root system voxel graph are connected based on the spatial positions of the current root system voxel grid and the adjacent root system voxel grids, and the weights of the root system voxel graph are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; The noise filtering module constructs a root system voxel spanning tree based on the weights of the root system voxel graph and the root system voxel graph according to the minimum spanning tree algorithm, and then determines whether there is noise in the root system voxel graph, and filters the edges corresponding to the noise to obtain the initial root system skeleton graph; The burr filtering module obtains the connection relationship between the current root system voxel grid and the adjacent root system voxel grid in the initial root system skeleton graph to obtain the root system connection data, and determines whether there are burrs in the initial root system skeleton graph, and filters the root system voxel grids corresponding to the burrs to obtain the root system skeleton graph.

[0014] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following method is implemented: Obtain the plant root system point cloud and perform point cloud division to form a set of root system voxel grids; Construct a root system voxel graph. Among them, the edges of the root system voxel graph are connected based on the spatial positions of the current root system voxel grid and the adjacent root system voxel grids, and the weights of the root system voxel graph are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; Based on the weights of the root system voxel map and the root system voxel map, construct a root system voxel spanning tree based on the minimum spanning tree algorithm, and then determine whether there is noise in the root system voxel map, and filter the edges corresponding to the noise to obtain an initial root system skeleton map; Obtain the connection relationship between the current root system voxel grid and the adjacent root system voxel grids in the initial root system skeleton map to obtain root system connection data, and determine whether there are burrs in the initial root system skeleton map, and filter the root system voxel grids corresponding to the burrs to obtain the root system skeleton map.

[0015] A root system point cloud skeleton extraction device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the following method is implemented: Obtain the plant root system point cloud and perform point cloud division to form a root system voxel grid set; Construct a root system voxel map. Among them, the edges of the root system voxel map are obtained by connecting based on the spatial positions of the current root system voxel grid and the adjacent root system voxel grids, and the weights of the root system voxel map are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; Based on the weights of the root system voxel map and the root system voxel map, construct a root system voxel spanning tree based on the minimum spanning tree algorithm, and then determine whether there is noise in the root system voxel map, and filter the edges corresponding to the noise to obtain an initial root system skeleton map; Obtain the connection relationship between the current root system voxel grid and the adjacent root system voxel grids in the initial root system skeleton map to obtain root system connection data, and determine whether there are burrs in the initial root system skeleton map, and filter the root system voxel grids corresponding to the burrs to obtain the root system skeleton map.

[0016] Due to the adoption of the above technical solutions, the present invention has remarkable technical effects: The present invention divides the plant root system point cloud and constructs a root system voxel map. Based on the edges and weights of the root system voxel map, filters the noise existing in the root system voxels to realize the extraction of the root system skeleton map. The method of the present invention solves the problem of many misjudgments in the existing root system point cloud skeleton extraction methods. At the same time, the present invention filters the noise to solve the root system misjudgments caused by problems such as uneven distribution of plant root system point clouds, trailing shadows on the surface of plant root systems, and lack of smoothness. Brief Description of the Drawings

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 is a schematic flowchart of the method of the present invention; Figure 2 is a schematic overall diagram of the system of the present invention; Figure 3 is a schematic diagram of the plant root point cloud and the root voxel grid of the present invention; Figure 4 is a schematic diagram of the edges of the noise and the corresponding root voxel map in the present invention; Figure 5 is a schematic diagram of the burrs of the initial root skeleton map of the present invention. Specific Embodiments

[0019] The following further elaborates on the present invention in conjunction with embodiments. The following embodiments are explanations of the present invention, and the present invention is not limited to the following embodiments.

[0020] Embodiment 1: A method for extracting the root point cloud skeleton, as Figure 1 shown, includes the following steps: S100. Obtain the plant root point cloud and perform point cloud partitioning to form a set of root voxel grids; S200. Construct a root voxel map. Among them, the edges of the root voxel map are obtained by connecting based on the spatial positions of the current root voxel grid and the adjacent root voxel grids, and the weights of the root voxel map are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; S300. Based on the weights and the root voxel map of the root voxel map, construct a root voxel spanning tree based on the minimum spanning tree algorithm, and then determine whether there is noise in the root voxel map, and filter the edges corresponding to the noise to obtain an initial root skeleton map; S400. Obtain the connection relationship between the current root voxel grid and the adjacent root voxel grids in the initial root skeleton map to obtain root connection data, and determine whether there are burrs in the initial root skeleton map, and filter the root voxel grids corresponding to the burrs to obtain a root skeleton map.

[0021] The present invention obtains the plant root point cloud, and forms a root voxel grid set through point cloud division. In the prior art, a voxel is a three-dimensional concept, commonly used in fields such as computer graphics, medical imaging, and 3D modeling. It is the basic unit in three-dimensional space, similar to pixels in two-dimensional images. A voxel can be defined as a cube or a more complex geometric shape, and is usually used to construct 3D models or represent 3D data, similar to pixels used to construct 2D images. Based on the root voxel grid set, a root voxel map is constructed. By combining the weights of the root voxel map, a root voxel spanning tree is obtained through the minimum spanning tree algorithm, and the noise in the root voxel map is judged. The edges of the root voxel map corresponding to the noise are filtered. Furthermore, through the connection relationship between root voxel points, the root voxel grids corresponding to the burrs in the initial root skeleton map are filtered to obtain the root skeleton map. Since the plant root point cloud obtained by the 3D reconstruction algorithm has problems such as uneven distribution, trailing shadows, burrs, and discrete noise points on the root surface, the noise and burrs in the plant root point cloud are filtered by the method of the present invention, improving the accuracy of the root skeleton map, and there is no need for fine adjustment of parameters, with stronger applicability and generalization ability.

[0022] In this embodiment, the plant root point cloud is a dense point cloud. The high-density plant root point cloud data obtained by high-precision 3D scanning technology can detailly reflect the morphology, structure, and spatial distribution of the roots. The plant root point cloud is composed of a large number of points with coordinate information in three-dimensional space, which can provide very detailed three-dimensional space information, clearly reflecting the details and geometric features of the object surface, and is usually obtained through technologies such as lidar, depth cameras, or photogrammetry.

[0023] The plant root point cloud in this embodiment has three-dimensional space boundaries, including horizontal boundaries, longitudinal boundaries, and vertical boundaries. Among them, the horizontal boundary is , the longitudinal boundary is , the vertical boundary is , and the preset grid threshold , the plant root point cloud within the three-dimensional space boundary is partitioned by presetting a grid threshold to obtain an initial root voxel grid set. The number of plant root point clouds in all the initial root voxel grids in the initial root voxel grid set is calculated to form a point cloud number set. A preset point cloud number threshold is set, and the point cloud number set is compared. The initial root voxels corresponding to the point cloud numbers in the point cloud number set that do not meet the point cloud number threshold are filtered to obtain a filtered voxel grid set. In this embodiment, based on the idea of the VoxelThinning algorithm, the filtered root voxel grid set is further refined and deleted. The filtered root voxels in the filtered root voxel grid set are deleted through a preset template. If deleting the current filtered root voxel grid does not affect the connectivity of the filtered root voxel grid set, then the current filtered root voxel grid is deleted; otherwise, it is retained. Through iterative deletion of the filtered root voxel grid set, a root voxel grid set is obtained. , where, represents the th root voxel grid, represents the number of root voxel grids. In the method of this embodiment , each root voxel grid is a cube, and the center of the cube is at the spatial position , and the side length, i.e., the grid threshold, is . A schematic diagram of the plant root point cloud and the root voxel grid is as shown in Figure 3 .

[0024] For all in the root voxel grid set , in this embodiment, within the range of 26 adjacent root voxel grids of , a search space for the current root voxel grid is constructed. Search within the search space to determine whether there is any other adjacent root voxel grid . If there is, connect the spatial position of the current root voxel grid to the spatial position of the adjacent root voxel grid to form an edge of the root voxel graph; with the spatial position of the current root voxel grid as the center of the sphere and the grid threshold as the diameter, construct a spherical point cloud search area. Within the point cloud search area, search for the number of point clouds in the plant root point cloud that fall within the point cloud search area, and calculate the point cloud density within the point cloud search area to obtain the root point cloud density . Among them, the specific process of calculating the root point cloud density is as follows: [IDX, ~]= rangesearch(PC.Location, voxelCenter, radius, 'Distance','euclidean','NSMethod','kdtree'); PC.Location: The three-dimensional spatial coordinates of the plant root point cloud ; voxelCenter: Each row is ; Radius: The radius of the point cloud search space, taken as GridSize / 2; IDX: The cell of, each cell contains The point cloud index within the spherical search range of ; Obtain the root point cloud density of all root voxel grids in the root voxel grid set, calculate the maximum value of all root point cloud densities, and through normalization, obtain the normalized point cloud density, expressed as follows:

[0025]

[0026] Based on the normalized point cloud density of the current root voxel grid and the normalized point cloud density of adjacent root voxel grids, calculate the edge The weight of the root voxel graph, expressed as follows:

[0027] Among them, Represents the maximum value of the root point cloud density, Represents the root point cloud density, Represents the number of root voxel grids, Represents the th normalized point cloud density, Represents the th root voxel grid, Represents the th root voxel grid and the th root voxel grid, the weight of the root voxel graph between them, , Represents the th normalized point cloud density.

[0028] After obtaining the edges and weights of the root voxel graph, construct the root voxel graph , expressed as follows:

[0029]

[0030]

[0031] Among them, represents the root voxel graph, represents the set of edges of the root voxel graph, represents the set of weights of the root voxel graph.

[0032] In this embodiment, the noise includes ghosting and noise points. The noise and the corresponding edges of the root voxel graph are as Figure 4 shown. Since the point cloud of the plant root system has a large point cloud density on the root system, while the point cloud density at the noise location is small, therefore, a root voxel graph is constructed such that the point cloud on the root system with a large density is more likely to be connected to form a root voxel graph. In the root voxel graph, it is necessary to filter the noise. Among them, filtering the noise in the root voxel graph to generate the edges of the root voxel graph to obtain the initial root skeleton graph includes the following steps: Step 1: Sort the edges of the corresponding root voxel graph according to the weights of the root voxel graph to obtain a sorted voxel graph; Step 2: Initialize the root voxel generation set , and determine whether the vertices corresponding to the edges of the sorted voxel graph belong to the root generation set. If not, add the edges of the sorted voxel graph to the root voxel generation set. If so, skip; Step 3: Iteratively traverse the edges of the sorted voxel graph until the edges in the root generation set can connect all the root voxel grids in the sorted voxel graph, and the sum of the weights of the root voxel generation set is the smallest, to obtain a root voxel generation tree; Step 4: Through the root voxel generation tree, judge the noise in the root voxel graph, and filter the edges in the root voxel graph that do not belong to the root voxel generation tree to obtain the initial root skeleton graph .

[0033] In this embodiment, filtering the edges of the root voxel graph is implemented by a minimum spanning tree. Specifically, the minspantree function in MATLAB for the minimum spanning tree is used. The pseudocode for this process is as follows: Gske = minspantree(Gvox,'Method','sparse') Among them, represents the root voxel graph, represents the initial root skeleton graph.

[0034] After filtering the edges of the root voxel graph corresponding to the ghosting and noise points in the root voxel graph, there are burrs in the initial root skeleton graph, asFigure 5 As shown in the figure, therefore, it is necessary to filter the root voxel grids corresponding to the burrs in the initial root system skeleton diagram to obtain the root system skeleton diagram, which includes the following steps: Step 1: Initialize the root connection data of the current root voxel grid to 0, traverse the edges of the initial root system skeleton diagram, and obtain the number of edges connected to the current root voxel grid to get the root connection data of the current root voxel grid; Step 2: Traverse the root voxel grids in the initial root system skeleton diagram to obtain the root connection data of all root voxel grids. In this embodiment, the root connection data is calculated by the degree function in MATLAB and is expressed as follows: Deg = degree( ) where, represents the initial root system skeleton diagram, represents the initial root system skeleton diagram the root connection data of the root voxel grid in; Step 3: Determine whether the first voxel data and the second voxel data meet the burr condition. In this embodiment, the burr condition is set as: the first voxel data is 1 and the second voxel data is greater than 2; Step 4: If the root connection data of the current root voxel grid is 1 and the root connection data of the adjacent root voxel grid is greater than 2, then determine that the current root voxel grid is a burr, and through iteration, filter the burrs in the initial root system skeleton diagram to obtain the root system skeleton diagram.

[0035] Through the root point cloud skeleton extraction method of the present invention, the influence of ghosting, discrete noise points and burrs in the plant root point cloud on the skeleton extraction can be eliminated, and the accuracy of the root system skeleton diagram can be improved.

[0036] Embodiment 2: A root point cloud skeleton extraction system, as Figure 2 shown, includes a point cloud division module 100, a voxel map construction module 200, a noise filtering module 300 and a burr filtering module 400; The point cloud division module 100 obtains the plant root point cloud and performs point cloud division to form a root voxel grid set; The voxel map construction module 200 constructs a root voxel map. Among them, the edges of the root voxel map are connected based on the spatial positions of the current root voxel grid and the adjacent root voxel grids, and the weights of the root voxel map are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; The noise filtering module 300 constructs a root voxel spanning tree based on the weights of the root voxel map and the root voxel map according to the minimum spanning tree algorithm, and then determines whether there is noise in the root voxel map, and filters the edges corresponding to the noise to obtain an initial root skeleton map; The burr filtering module 400 obtains the connection relationship between the current root voxel grid and the adjacent root voxel grids in the initial root skeleton map to obtain root connection data, and determines whether there are burrs in the initial root skeleton map, and filters the root voxel grids corresponding to the burrs to obtain a root skeleton map.

[0037] All changes and variations made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also fall within the scope of the present invention.

[0038] Each embodiment in this specification is described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same and similar parts among the embodiments, reference can be made to each other.

[0039] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, devices, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0040] The present invention is described with reference to the flowcharts and / or block diagrams of methods, terminal devices (systems), and computer program products according to the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, and the combination of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing terminal devices to generate a machine, so that the instructions executed by the processors of the computer or other programmable data processing terminal devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0041] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1The functions specified in one or more boxes.

[0042] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide for implementing the steps of the functions specified in one Figure 1 process or multiple processes and / or boxes Figure 1 step of the functions specified in one or more boxes.

[0043] It should be noted that: "An embodiment" or "embodiments" mentioned in the specification means that the specific features, structures or characteristics described in connection with the embodiments are included in at least one embodiment of the present invention. Therefore, the phrases "an embodiment" or "embodiments" appearing throughout the specification do not necessarily all refer to the same embodiment.

[0044] In addition, it should be noted that for the specific embodiments described in this specification, the shapes, names of the components, etc. can be different. Any equivalent or simple changes made according to the structure, features and principles described in the inventive concept of the present invention are included in the protection scope of the present invention. Those skilled in the art to which the present invention pertains can make various modifications or supplements to the described specific embodiments or use similar means for substitution, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should fall within the protection scope of the present invention.

Claims

1. A method for extracting the root system point cloud skeleton, characterized in that, It includes the following steps: Obtain the plant root point cloud and perform point cloud partitioning to form a root voxel grid set; Construct a root voxel graph. Among them, the edges of the root voxel graph are obtained by connecting based on the spatial positions of the current root voxel grid and adjacent root voxel grids, and the weights of the root voxel graph are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; Based on the weights of the root voxel graph and the root voxel graph, construct a root voxel spanning tree based on the minimum spanning tree algorithm, and then determine whether there is noise in the root voxel graph, and filter the edges corresponding to the noise to obtain an initial root skeleton graph; Obtain the connection relationship between the current root voxel grid and adjacent root voxel grids in the initial root skeleton graph to obtain root connection data, and determine whether there are burrs in the initial root skeleton graph, and filter the root voxel grids corresponding to the burrs to obtain a root skeleton graph.

2. The root point cloud skeleton extraction method according to claim 1, wherein The point cloud partitioning of the plant root point cloud includes the following steps: Obtain the plant root point cloud and determine the three-dimensional space boundary of the plant root point cloud. Among them, the three-dimensional space boundary includes a horizontal boundary, a vertical boundary, and a vertical boundary; Perform point cloud partitioning on the plant root point cloud within the three-dimensional space boundary according to a preset grid threshold to obtain an initial root voxel grid set; Obtain the number of point clouds contained in all initial root voxel grids in the initial root voxel grid set to obtain a point cloud quantity set, set a point cloud quantity threshold, obtain the number of point clouds in the point cloud quantity set that do not meet the point cloud quantity threshold, and filter the corresponding initial root voxel grids to obtain a filtered voxel grid set; In the filtered voxel grid set, if the filtered voxel grid set still has connectivity after deleting the current filtered voxel grid, then delete the current filtered voxel grid, and obtain a root voxel grid set through iterative deletion.

3. The root point cloud skeleton extraction method according to claim 1, characterized in that The step of connecting based on the spatial positions of the current root voxel grid and adjacent root voxel grids to obtain the edges of the root voxel graph includes the following steps: Obtain the spatial position of the root voxel grid through the center position of the root voxel grid; Preset a search threshold, construct a search space for the current root voxel grid based on the search threshold, and determine whether there are adjacent root voxel grids within the search space; If there are, connect the spatial positions of the current root voxel grid and adjacent root voxel grids to obtain the edges of the root voxel graph.

4. The root point cloud skeleton extraction method according to claim 1, characterized in that The step of obtaining the weights of the root voxel graph by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space includes the following steps: Take the spatial position of the root voxel grid as the center of the sphere and the preset grid threshold as the diameter to construct a point cloud search area; Obtain the number of plant root point clouds within the point cloud search area and perform density calculation to obtain the root point cloud density; Perform normalization processing on the root point cloud density to obtain a normalized point cloud density; Calculate based on adjacent normalized point cloud densities to obtain the weights of the root voxel graph; Among them, the normalized point cloud density and the weights of the root voxel graph are expressed as follows: Among them, , represents the maximum value of the root system point cloud density, represents the root system point cloud density, represents the number of root system voxel grids, represents the th normalized point cloud density, represents the th root system voxel grid, represents the th root system voxel grid and the th root system voxel grid, the weight of the root system voxel map between them, represents the th normalized point cloud density.

5. The root point cloud skeleton extraction method according to claim 1, characterized in that Based on the weights of the root voxel map and the root voxel map, construct a root voxel spanning tree using the minimum spanning tree algorithm, and then obtain the noise in the root voxel map, and filter the edges corresponding to the noise to obtain an initial root skeleton map, including the following steps: Sort the root voxel map according to the weights of the root voxel map to obtain a sorted voxel map; Initialize the root generation set, and determine whether the root voxel grid corresponding to the edge of the current sorted voxel map belongs to the root generation set. If not, add the edge of the current sorted voxel map to the root generation set; If it belongs, filter the edge of the current sorted voxel map, and traverse the sorted voxel map until the edges of the root generation set can connect all the root voxel grids in the sorted voxel map and the sum of the weights of the root voxel generation set is the smallest, to obtain the root voxel spanning tree; Filter the edges in the root voxel map that do not belong to the root voxel spanning tree to obtain an initial root skeleton map.

6. The method for extracting the root system point cloud skeleton according to claim 1, wherein The obtaining of the connection relationship between the current root voxel grid and adjacent root voxel grids to obtain root connection data includes the following steps: Initialize the root connection data of the current root voxel grid in the initial root skeleton map to 0; Traverse the edges of the initial root skeleton map, obtain the number of edges of the initial root skeleton map connected to the current root voxel grid, and obtain the root connection data corresponding to all root voxel grids.

7. The root system point cloud skeleton extraction method according to claim 1, characterized in that The judging whether there are burrs in the initial root skeleton map and filtering the root voxel grids corresponding to the burrs to obtain a root skeleton map includes the following steps: Obtain the root connection data of the current root voxel grid in the initial root skeleton map to obtain the first voxel data, and obtain the root connection data of the adjacent root voxel grid to obtain the second voxel data; Based on the first voxel data and the second voxel data, judge whether the root voxel grid meets the burr condition. If it meets, it is judged as a burr, where the burr condition is at least: the first voxel data is 1 and the second voxel data is greater than 2; Traverse the initial root skeleton map, and filter the root voxel grids corresponding to the burrs to obtain a root skeleton map.

8. A root system point cloud skeleton extraction system, characterized in that, It includes a point cloud division module, a voxel map construction module, a noise filtering module and a burr filtering module; The point cloud division module obtains the plant root point cloud and performs point cloud division to form a root voxel grid set; The voxel map construction module constructs a root voxel map, where the edges of the root voxel map are connected based on the spatial positions of the current root voxel grid and adjacent root voxel grids, and the weights of the root voxel map are obtained by constructing a point cloud search space and obtaining the point cloud density within the point cloud search space; The noise filtering module, based on the weights of the root voxel map and the root voxel map, constructs a root voxel spanning tree using the minimum spanning tree algorithm, and then judges whether there is noise in the root voxel map, and filters the edges corresponding to the noise to obtain an initial root skeleton map; The burr filtering module obtains the connection relationship between the current root voxel grid and adjacent root voxel grids in the initial root skeleton map to obtain root connection data, and judges whether there are burrs in the initial root skeleton map, and filters the root voxel grids corresponding to the burrs to obtain a root skeleton map.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method according to any one of claims 1 to 7.

10. A root system point cloud skeleton extraction device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, it implements the method according to any one of claims 1 to 7.

Images