A method and system for extracting features from unordered 3D discrete point clouds using a slice-based approach.

Abstract

This invention discloses a method and system for extracting features from disordered 3D discrete point clouds using a slice-based approach. The method includes: obtaining the effective range and density of disordered 3D discrete points, and setting slice intervals in the X, Y, and Z directions according to the density; slicing the disordered 3D discrete points in the X, Y, and Z directions from smallest to largest according to the set slice intervals, obtaining two-dimensional slices in each direction; calculating the concave hull of each slice in the X, Y, and Z directions based on the two-dimensional slice data in each direction, and marking the points contained within the outer boundary of the concave hull; counting the repeatedly marked points to form a set of repeatedly marked points; and performing voxel filtering on the set of repeatedly marked points to obtain the extracted feature points. Inspired by medical CT slices, this invention can handle 3D entities of arbitrary shape and size, and can solve the problems of complexity and time-consuming processing in existing disordered 3D discrete point cloud feature extraction algorithms.

Description

Technical Field

[0001] This invention belongs to the field of laser point cloud feature extraction, and particularly relates to a method for extracting feature points from three-dimensional discrete point clouds by continuously slicing them in the x, y, and z directions to obtain and mark the concave hulls. Specifically, it is a method and system for extracting feature points from disordered three-dimensional discrete point clouds by slicing them in the x, y, and z directions. Background Technology

[0002] With the rapid development of 3D acquisition technology, the acquisition of disordered 3D discrete point cloud data has become increasingly convenient. Disordered 3D discrete point clouds are a fundamental data source for 3D spatial information processing; however, due to limitations in the acquisition environment, methods, and sensor positioning, data acquired multiple times often cannot be directly stitched together. Therefore, point cloud registration (PCR) technology has emerged as a key to achieving accurate data alignment. In point cloud data processing tasks, feature extraction, as an important preprocessing step, provides crucial support for subsequent feature matching, registration, and overall alignment; the two are closely related. Efficient and accurate feature extraction is the foundation of registration quality. Existing point feature extraction methods mainly include SIFT3D, Harris3D, and NARF. While these methods can extract point features well, they all share a common drawback: their algorithms are relatively complex, time-consuming, and have specific requirements for the point cloud, limiting the rapid processing of 3D geographic information data. Summary of the Invention

[0003] To overcome the shortcomings of the prior art, this invention provides a method and system for extracting features from disordered three-dimensional discrete point clouds in a slice-like manner. Inspired by medical CT slices, it can process three-dimensional entities of arbitrary shape and size, and can solve the problems of complexity and time-consuming processing of existing disordered three-dimensional discrete point cloud feature extraction algorithms.

[0004] According to one aspect of the present invention, a method for extracting slice-type features from disordered three-dimensional discrete point clouds is provided, comprising:

[0005] Obtain the effective range and density of disordered three-dimensional discrete points, and set the slicing interval in the X, Y, and Z directions according to the density;

[0006] Based on the set slicing interval, the disordered three-dimensional discrete points are sliced ​​in the X, Y, and Z directions from small to large to obtain two-dimensional slices in each direction.

[0007] Based on the two-dimensional slice data in each direction, calculate the concave hull of each slice in the X, Y, and Z directions, and mark the points contained in the outer boundary of the concave hull;

[0008] Count the points that are marked repeatedly to form a set of points with repeated markings;

[0009] Voxel filtering is performed on the set of repeatedly labeled points to obtain the extracted feature points.

[0010] As a further technical solution, the disordered three-dimensional discrete points are sliced ​​in the X, Y, and Z directions from smallest to largest, including:

[0011] Statistically calculate the maximum and minimum values ​​of an unordered three-dimensional discrete point in the X, Y, and Z directions;

[0012] Using the intervals dx, dy, and dz in the corresponding X, Y, and Z directions, the disordered three-dimensional discrete points are sliced ​​from smallest to largest.

[0013] As a further technical solution, two-dimensional slices are obtained in various directions, including:

[0014] A planar coordinate system is established for the cut surface using two coordinate axes perpendicular to the slice direction. The slice direction coordinates are set to 0 to reduce the 3D point cloud to a 2D point cloud.

[0015] As a further technical solution, the concave hulls of each tangent in the X, Y, and Z directions are calculated respectively, and the points contained within the outer boundary of the concave hull are marked, including:

[0016] The concave hull of each cross-section is determined using the classic rolling ball method.

[0017] For the points on the obtained concave hull, their indices in the original unordered 3D discrete point cloud are found using a KD tree and then marked.

[0018] As a further technical solution, the concave hull of each cross-section is determined using the classical rolling ball method, including:

[0019] Find the vertex P0 of a two-dimensional point set;

[0020] Find the leftmost point and use it as the first point P1 of the concave edge;

[0021] Using the sphere radius set by the rolling ball method, with P1 as the center, the set of points Pnearest within a fixed radius is selected.

[0022] Iterate through Pnearest, using the angle from P0 to P1 as a reference, rotate counterclockwise to find the point with the smallest counterclockwise angle and add it to the concave hull.

[0023] Repeat the above steps until no new points are found or a point that has already been added to the hull is encountered.

[0024] As a further technical solution, voxel filtering is performed on the repeatedly labeled point set, including:

[0025] The set of repeatedly labeled points is voxelized by voxel radius to obtain a set of feature points with consistent resolution and uniform distribution.

[0026] According to one aspect of the present invention, a system for extracting slicing features from disordered three-dimensional discrete point clouds is provided, comprising:

[0027] The first main module is used to obtain the effective range and density of disordered three-dimensional discrete points, and to set the slicing interval in the X, Y and Z directions according to the density.

[0028] The second main module is used to slice the disordered three-dimensional discrete points in the X, Y, and Z directions from small to large according to the set slicing interval, so as to obtain two-dimensional slices in each direction.

[0029] The third main module is used to calculate the concave hulls of each sectional plane in the X, Y, and Z directions based on the two-dimensional slice data in each direction, and to mark the points contained in the outer boundary of the concave hull.

[0030] The fourth main module is used to count the points that are marked repeatedly and form a set of points marked repeatedly.

[0031] The fifth main module is used to perform voxel filtering on the repeatedly labeled point set to obtain the extracted feature points.

[0032] According to one aspect of the present invention, an electronic device is provided, including a processor and a memory; the memory stores program instructions that are executed by the processor, and the processor invokes the program instructions to perform the steps of the described method for extracting sliced ​​features from disordered three-dimensional discrete point clouds.

[0033] According to one aspect of the present invention, a non-transitory computer read storage medium is provided, the non-transitory computer read storage medium storing computer instructions that cause the computer to perform the steps of the described disordered three-dimensional discrete point cloud slice feature extraction method.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] (1) Inspired by CT slices in medicine, this invention slices the point cloud and marks possible feature points based on the edges of the smallest concave polygons on the slices. Feature points are confirmed by repeated marking, and finally, voxel filtering is used to avoid dense feature points. Compared with some classic feature point extraction algorithms such as SIFT, Harris3D, and NARF, the method of this invention is simpler and more direct in algorithm implementation. It does not require calculating the gradient and normal direction of the point cloud data, but only determines the feature points by statistically analyzing the concave hull edge points of each slice, resulting in higher processing efficiency.

[0036] (2) The advantage of the method of this invention lies in its simple calculation, without relying on more complex geometric properties such as calculating derivatives. Therefore, it is particularly suitable for embedded devices with limited computing power but requiring real-time and efficient processing of point cloud features. For example, low-power microcontrollers, unmanned vehicles, robots, or DSP systems in the field of industrial control. These lightweight devices have higher requirements for algorithm complexity in point cloud preprocessing. This method achieves feature point recognition through extremely simple calculations, maximizing processing speed while ensuring the necessary feature quality, which is why it has a strong advantage in application and promotion. It fills the technical gap in feature extraction for real-time low-power devices. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating a method for extracting features from unordered three-dimensional discrete point clouds using a slice-based approach, as provided in an embodiment of the present invention.

[0039] Figure 2 (a)-(c) are schematic diagrams of the slicing principle provided in the embodiments of the present invention.

[0040] Figure 3 This is a schematic diagram of the rolling ball method for finding the concave hull provided in an embodiment of the present invention.

[0041] Figure 4 This is a schematic diagram illustrating an example of feature extraction results provided in an embodiment of the present invention.

[0042] Figure 5 This is a schematic diagram of the structure of a disordered three-dimensional discrete point cloud slice-type feature extraction system provided in an embodiment of the present invention. Detailed Implementation

[0043] To address the limitations of existing algorithms, which are complex, time-consuming, and require specific point clouds, hindering the rapid processing of 3D geographic information data, this invention, inspired by CT slices in medicine, proposes a slice-based feature extraction algorithm for unordered 3D discrete point clouds. CT slices can decompose any entity into sheet-like structures. The proposed algorithm is similar to this slicing process. First, it performs slicing operations and dimensionality reduction on the x, y, and z directions of the unordered 3D discrete points. Then, it finds the smallest concave polygon on the slice and marks the edge points. Finally, it performs voxel filtering on the repeatedly marked point set to obtain the final feature points. This algorithm is simple to process, computationally efficient, has no requirements on data points, and can handle 3D entities of arbitrary shapes and sizes, making it particularly suitable for rapid feature extraction from 3D geographic information data. This method avoids complex 3D spatial feature calculations. The computation within each slice is simple, and the numerous independent subtasks formed by the slices facilitate parallel processing, enabling stable description of the point cloud.

[0044] The terms “comprising” and “having”, and any variations thereof, in the specification, claims, and accompanying drawings of this invention are intended to cover a non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form new technical solutions. Such combinations are not bound by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0046] Inspired by CT scans in everyday medical practice, this invention proposes a slice-based feature extraction method for disordered 3D discrete point clouds. The core of the method comprises three processes: first, slicing the disordered 3D discrete points; second, finding the smallest concave polygon on the slice and marking its edge points; and finally, performing voxel filtering on the repeatedly marked point set to obtain the final feature points. This method is simple to process, avoids complex 3D spatial feature calculations, has high computational efficiency, is not limited by the number of data points, can handle 3D entities of arbitrary shape and size, and the numerous independent subtasks formed by slicing facilitate parallel processing, making it particularly suitable for rapid feature extraction from 3D geographic information data.

[0047] like Figure 1 As shown, an embodiment of the present invention provides a method for extracting features from unordered three-dimensional discrete point clouds in a slice-like manner, comprising the following steps:

[0048] Step 1: Obtain the effective range of the unordered three-dimensional discrete points, and in... X,Y,Z Set appropriate density values ​​for each of the three directions. dx, dy, dz As intervals, in the corresponding directions (e.g.) X The unordered three-dimensional discrete points are sliced ​​from smallest to largest. The schematic diagram of the slicing principle is as follows: Figure 2 As shown in (a)-(c).

[0049] Step 1.1: Calculate the maximum and minimum values ​​of the three-dimensional discrete points in the X, Y, and Z directions. The specific process is as follows:

[0050] (1) Calculate the maximum and minimum values ​​minx, minY, minZ, maxX, maxY, maxZ in the X, Y and Z directions respectively.

[0051] (2) Construct a KD tree and use existing algorithms such as octree statistical method and grid statistical method to calculate point cloud density.

[0052] (3) Using dx, dy, and dz as intervals, slice the disordered three-dimensional discrete points in the X, Y, and Z directions from smallest to largest.

[0053] in, dx, dy, dz The calculation formula is as follows:

[0054]

[0055] Step 2: Reduce the dimensionality of the 3D slices according to the slice direction so that the points in the slices are located on the same plane.

[0056] Step 2.1: Establish a planar coordinate system for the cut surface using two coordinate axes (such as the Y and Z axes) perpendicular to the slice direction, and set the slice direction coordinates to 0. This reduces the 3D point cloud to a 2D point cloud, considering only the planar coordinates of the slice.

[0057] Step 3: For all 2D slice data in sequence, set a threshold and find the concave hull (i.e., the smallest concave polygon containing all points) on its slice surface. Mark the points contained in the outer boundary of the smallest concave polygon, such as... Figure 3 As shown.

[0058] Step 3.1: Use the classic rolling ball method to find the concave hull of each cross-section. The specific process is as follows:

[0059] (1) Find the pole P0 of the two-dimensional point set;

[0060] (2) Find the leftmost point, and use this point as the first point P1 of the concave edge;

[0061] (3) The radius of the ball is set by the rolling ball method, and the set of points P within a fixed radius range is selected with P1 as the center. nearest ;

[0062] (4) Traverse P nearest Using the P0->P1 angle as a reference, rotate counterclockwise to find the point with the smallest counterclockwise angle and add it to the concave hull;

[0063] (5) Repeat the above steps until no new points are found or a point that has been added to the hull is encountered.

[0064] Step 3.2: For the points on the concave hull obtained in Step 3.1, find their indices in the original unordered 3D discrete point cloud using a KD tree and mark them.

[0065] Step 4: Process the Y and Z directions according to the above steps, and count the points that are marked three times to obtain the set of points marked repeatedly.

[0066] Step 4.1: Using a counter, in the slices along the X, Y, and Z directions, the counter of each marked point is incremented by 1. A point that is marked repeatedly is defined as being marked in all three directions (X, Y, and Z), i.e., its counter is 3. In particular, the set of points marked repeatedly can be obtained.

[0067] Step 5: Perform voxel filtering on the repeatedly labeled point set to obtain the final feature points. For example... Figure 4 As shown.

[0068] Step 5.1: Voxelize the repeatedly marked point set with voxel radius d to obtain a set of feature points with consistent resolution and uniform distribution. The point cloud within the voxel is downsampled to 1, which can effectively avoid the situation of dense feature points.

[0069] After completing the above five steps, the feature extraction of the unordered 3D discrete point cloud can be achieved. This process is very similar to CT slicing in medicine. The point cloud is sliced, and potential feature points are marked based on the edges of the smallest concave polygons on the slices. Feature points are confirmed through repeated marking, and finally, voxel filtering is used to avoid dense feature point clusters. Compared with some classic feature point extraction algorithms such as SIFT, Harris3D, and NARF, this method is simpler and more direct in its implementation. It does not require calculating the gradient and normal direction of the point cloud data; it only determines feature points by statistically analyzing the concave hull edge points of each slice, resulting in higher processing efficiency.

[0070] In practice, the above process can be automated using computer software technology.

[0071] The implementation of the various embodiments of the present invention is based on programmed processing through a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of the present invention are encapsulated into various modules. Based on this reality, and building upon the above embodiments, the embodiments of the present invention provide a disordered three-dimensional discrete point cloud slice-based feature extraction system, which is used to execute a disordered three-dimensional discrete point cloud slice-based feature extraction method from the above method embodiments.

[0072] See Figure 5 The system includes: a first main module for acquiring the effective range and density of disordered three-dimensional discrete points, and setting the slicing interval in the X, Y, and Z directions according to the density; a second main module for slicing the disordered three-dimensional discrete points in the X, Y, and Z directions according to the set slicing interval, from small to large, to obtain two-dimensional slices in each direction; a third main module for calculating the concave hull of each slice in the X, Y, and Z directions based on the two-dimensional slice data in each direction, and marking the points contained in the outer boundary of the concave hull; a fourth main module for counting the repeatedly marked points to form a set of repeatedly marked points; and a fifth main module for performing voxel filtering on the set of repeatedly marked points to obtain the extracted feature points.

[0073] This invention provides a slice-based feature extraction system for disordered 3D discrete point clouds, addressing the problems of complexity and time-consuming processing in existing disordered 3D discrete point cloud feature extraction algorithms. Figure 5 Several modules in the process perform slicing operations and dimensionality reduction on the x, y, and z directions of the disordered 3D discrete points. The minimum concave polygon is found on the slice and the edge points are marked. Finally, the set of points marked multiple times is subjected to voxel filtering to obtain the final feature points.

[0074] It should be noted that the system embodiments provided by this invention, in addition to implementing the methods in the above method embodiments, are also used to implement the methods in other method embodiments provided by this invention. The difference lies only in setting corresponding functional modules, and their principles are basically the same as those of the above system embodiments provided by this invention. As long as those skilled in the art, based on the above system embodiments and referring to the specific technical solutions in other method embodiments, obtain corresponding technical means and technical solutions composed of these technical means by combining technical features, and improve the modules in the above system embodiments while ensuring the practicality of the technical solutions, they can obtain corresponding system-like embodiments for implementing the methods in other method-like embodiments. For example:

[0075] Based on the above system embodiments, as a preferred embodiment, the unordered three-dimensional discrete point cloud slice-type feature extraction system provided in this embodiment of the invention, wherein the second main module is further configured to execute the following instructions:

[0076] Statistically calculate the maximum and minimum values ​​of an unordered three-dimensional discrete point in the X, Y, and Z directions;

[0077] Using the intervals dx, dy, and dz in the corresponding X, Y, and Z directions, the disordered three-dimensional discrete points are sliced ​​from smallest to largest.

[0078] Based on the above system embodiments, as a preferred embodiment, the unordered three-dimensional discrete point cloud slice-type feature extraction system provided in this embodiment of the invention, wherein the second main module is further configured to execute the following instructions:

[0079] A planar coordinate system is established for the cut surface using two coordinate axes perpendicular to the slice direction. The slice direction coordinates are set to 0 to reduce the 3D point cloud to a 2D point cloud.

[0080] Based on the above system embodiments, as a preferred embodiment, the unordered three-dimensional discrete point cloud slice-type feature extraction system provided in this embodiment of the invention, wherein the third main module is further configured to execute the following instructions:

[0081] The concave hull of each cross-section is determined using the classic rolling ball method.

[0082] For the points on the obtained concave hull, their indices in the original unordered 3D discrete point cloud are found using a KD tree and then marked.

[0083] Based on the above system embodiments, as a preferred embodiment, the unordered three-dimensional discrete point cloud slice-type feature extraction system provided in this embodiment of the invention, wherein the third main module is further configured to execute the following instructions:

[0084] Find the vertex P0 of a two-dimensional point set;

[0085] Find the leftmost point and use it as the first point P1 of the concave edge;

[0086] Using the sphere radius set by the rolling ball method, with P1 as the center, the set of points Pnearest within a fixed radius is selected.

[0087] Iterate through Pnearest, using the angle from P0 to P1 as a reference, rotate counterclockwise to find the point with the smallest counterclockwise angle and add it to the concave hull.

[0088] Repeat the above steps until no new points are found or a point that has already been added to the hull is encountered.

[0089] Based on the above system embodiments, as a preferred embodiment, the unordered three-dimensional discrete point cloud slice feature extraction system provided in this embodiment of the invention, wherein the fifth main module is further configured to execute the following instructions:

[0090] The set of repeatedly labeled points is voxelized by voxel radius to obtain a set of feature points with consistent resolution and uniform distribution.

[0091] Based on the same inventive concept as the above embodiments, the present invention also provides an electronic device, including a processor and a memory; the memory stores program instructions executed by the processor, and the processor calls the program instructions to perform the following steps of the disordered three-dimensional discrete point cloud slice feature extraction method:

[0092] Obtain the effective range and density of disordered three-dimensional discrete points, and set the slicing interval in the X, Y, and Z directions according to the density;

[0093] Based on the set slicing interval, the disordered three-dimensional discrete points are sliced ​​in the X, Y, and Z directions from small to large to obtain two-dimensional slices in each direction.

[0094] Based on the two-dimensional slice data in each direction, calculate the concave hull of each slice in the X, Y, and Z directions, and mark the points contained in the outer boundary of the concave hull;

[0095] Count the points that are marked repeatedly to form a set of points with repeated markings;

[0096] Voxel filtering is performed on the set of repeatedly labeled points to obtain the extracted feature points.

[0097] Based on the same inventive concept as the above embodiments, the present invention also provides a non-transitory computer read storage medium, wherein the non-transitory computer read storage medium stores computer instructions, the computer instructions causing the computer to execute the steps of the disordered three-dimensional discrete point cloud slice feature extraction method described above:

[0098] Obtain the effective range and density of disordered three-dimensional discrete points, and set the slicing interval in the X, Y, and Z directions according to the density;

[0099] Based on the set slicing interval, the disordered three-dimensional discrete points are sliced ​​in the X, Y, and Z directions from small to large to obtain two-dimensional slices in each direction.

[0100] Based on the two-dimensional slice data in each direction, calculate the concave hull of each slice in the X, Y, and Z directions, and mark the points contained in the outer boundary of the concave hull;

[0101] Count the points that are marked repeatedly to form a set of points with repeated markings;

[0102] Voxel filtering is performed on the set of repeatedly labeled points to obtain the extracted feature points.

[0103] In summary, inspired by CT slices in medicine, this invention proposes a slice-based feature extraction algorithm for disordered 3D discrete point clouds. First, slicing and dimensionality reduction are performed on the x, y, and z directions of the disordered 3D discrete points. Then, the smallest concave polygon is found on the slices, and edge points are marked. Finally, voxel filtering is applied to the repeatedly marked point set to obtain the final feature points. This invention's method is simple to process, computationally efficient, and has no requirements on the number of data points. It can handle 3D entities of arbitrary shape and size, and is particularly suitable for rapid feature extraction from 3D geographic information data. Furthermore, this invention avoids complex 3D spatial feature calculations. The computation within each slice is simple, and the numerous independent subtasks formed by the slices facilitate parallel processing, resulting in a stable point cloud description.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting features from unordered three-dimensional discrete point clouds in a slice-based manner, characterized in that, include: To obtain the effective range and density of disordered 3D discrete points, and to set the slicing intervals in the X, Y, and Z directions according to the density, the following steps are taken: 1) Calculate the maximum and minimum values ​​of the 3D discrete points in the X, Y, and Z directions, respectively: minx, minY, minZ, maxX, maxY, maxZ; 2) Construct a KD tree and calculate the point cloud density using octree or grid statistical methods; 3) Slice the disordered 3D discrete points in the corresponding X, Y, and Z directions using dx, dy, and dz intervals, from smallest to largest. dx, dy, dz The calculation formula is as follows: ； Based on the set slicing interval, the disordered three-dimensional discrete points are sliced ​​in the X, Y, and Z directions from small to large to obtain two-dimensional slices in each direction. Based on the two-dimensional slice data in each direction, calculate the concave hull of each slice in the X, Y, and Z directions, and mark the points contained in the outer boundary of the concave hull; Count the points that are marked repeatedly to form a set of points marked repeatedly; Voxel filtering is performed on the set of repeatedly labeled points to obtain the extracted feature points.

2. The method for extracting features from unordered three-dimensional discrete point clouds in a slice-like manner according to claim 1, characterized in that, Obtain two-dimensional slices in various directions, including: A planar coordinate system is established for the cut surface using two coordinate axes perpendicular to the slice direction. The slice direction coordinates are set to 0 to reduce the 3D point cloud to a 2D point cloud.

3. The method for extracting features from unordered three-dimensional discrete point clouds in a slice-like manner according to claim 1, characterized in that, Calculate the concave hull of each tangent plane in the X, Y, and Z directions, and mark the points contained within the outer boundary of the concave hull, including: The concave hull of each cross-section is determined using the classic rolling ball method. For the points on the obtained concave hull, their indices in the original unordered 3D discrete point cloud are found using a KD tree and then marked.

4. The method for extracting features from unordered three-dimensional discrete point clouds in a slice-like manner according to claim 3, characterized in that, The concave hull of each cross-section is determined using the classic rolling ball method, including: Find the vertex P0 of a two-dimensional point set; Find the leftmost point and use it as the first point P1 of the concave edge; Using the rolling ball method to set the radius of the ball, with P1 as the center, a set of points P within a fixed radius range is selected. nearest ; Traversing P nearest Using the angle P0->P1 as a reference, rotate counterclockwise to find the point with the smallest counterclockwise angle and add it to the concave hull. Repeat the above steps until no new points are found or a point that has already been added to the hull is encountered.

5. The method for extracting features from unordered three-dimensional discrete point clouds in a slice-like manner according to claim 1, characterized in that, Voxel filtering is performed on the set of repeatedly labeled points, including: The set of repeatedly labeled points is voxelized by voxel radius to obtain a set of feature points with consistent resolution and uniform distribution.

6. A slicing feature extraction system for disordered three-dimensional discrete point clouds, characterized in that, include: The first main module is used to obtain the effective range and density of unordered 3D discrete points, and to set the slicing intervals in the X, Y, and Z directions according to the density. This includes: calculating the maximum and minimum values ​​of the 3D discrete points in the X, Y, and Z directions, the process of which is as follows: 1) Calculating the maximum and minimum values ​​minx, minY, minZ, maxX, maxY, and maxZ in the X, Y, and Z directions respectively; 2) Constructing a KD tree and calculating the point cloud density using octree statistics or grid statistics; 3) Slicing the unordered 3D discrete points in the corresponding X, Y, and Z directions using dx, dy, and dz as intervals, from smallest to largest. dx, dy, dz The calculation formula is as follows: ； The second main module is used to slice the disordered three-dimensional discrete points in the X, Y, and Z directions from small to large according to the set slicing interval, so as to obtain two-dimensional slices in each direction. The third main module is used to calculate the concave hulls of each sectional plane in the X, Y, and Z directions based on the two-dimensional slice data in each direction, and to mark the points contained in the outer boundary of the concave hull. The fourth main module is used to count the points that are marked repeatedly and form a set of points marked repeatedly. The fifth main module is used to perform voxel filtering on the repeatedly labeled point set to obtain the extracted feature points.

7. An electronic device, characterized in that, It includes a processor and a memory; the memory stores program instructions that are executed by the processor, and the processor invokes the program instructions to perform the steps of the unordered three-dimensional discrete point cloud slice feature extraction method according to any one of claims 1 to 5.

8. A non-transitory computer read storage medium, characterized in that, The non-transitory computer read storage medium stores computer instructions, which cause the computer to perform the steps of the disordered three-dimensional discrete point cloud slice feature extraction method according to any one of claims 1 to 5.

Images