Point cloud-based corner point recognition method and system

Abstract

The application relates to a corner point recognition method and system based on a point cloud, which comprises the following steps: acquiring a surface point cloud; acquiring a boundary point set based on the surface point cloud; for each boundary point in the boundary point set, extracting a three-dimensional neighborhood point set of the boundary point in the surface point cloud, and obtaining a two-dimensional neighborhood point set of the boundary point through dimension reduction processing; constructing a first feature line according to the farthest point pairs in the two-dimensional neighborhood point set; identifying a corner region point set through the distribution characteristics of the points located on both sides of the first feature line; identifying a candidate corner point based on the farthest point pairs in the corner region point set and the farthest points of a second feature line constructed from the farthest point pairs; constructing a third feature line based on the two non-candidate corner points in the farthest point pairs and the farthest points of the corner region point set corresponding to each candidate corner point, and determining the density index and the cutoff distance of the candidate corner point; and determining the final corner point according to the density index and the cutoff distance of all candidate corner points. The application can accurately and quickly recognize corner points.

Description

Technical Field

[0001] This application relates to the field of point cloud processing technology, and in particular to a corner recognition method and system based on point clouds. Background Technology

[0002] Point clouds, as a direct record of the surface morphology of objects in three-dimensional space, have wide applications in reverse engineering, autonomous driving, robot navigation, digital preservation of cultural relics, and industrial inspection. Accurately and efficiently extracting the contour features of target objects from point clouds, especially the corner points, is a crucial step in high-level tasks such as 3D model reconstruction, shape analysis, and pose estimation. Corner points carry key geometric and topological information about the target object; the accuracy of corner point recognition directly affects the overall accuracy of point cloud processing, and the efficiency of corner point recognition directly affects the overall efficiency of point cloud processing.

[0003] However, existing point cloud-based corner recognition methods are prone to missing true corners in low-density areas of the point cloud and misidentifying false corners in flat, curved areas of the point cloud, resulting in low accuracy of corner recognition.

[0004] Therefore, there is a need to provide a corner recognition method based on point clouds to accurately and quickly identify corners. Summary of the Invention

[0005] This application provides a corner recognition method based on point clouds, which can accurately and quickly identify corners.

[0006] Firstly, this application provides a corner recognition method based on point clouds, including: Boundary point set acquisition steps: acquire the surface point cloud of a specific surface of the target object; based on the surface point cloud, acquire the boundary point set used to describe the contour of the specific surface; Steps for obtaining a two-dimensional neighborhood point set: For each boundary point in the boundary point set, extract the three-dimensional neighborhood point set of the boundary point in the surface point cloud, and obtain the two-dimensional neighborhood point set of the boundary point through dimensionality reduction processing; Corner region point set acquisition steps: Construct a first feature line based on the farthest point pair in the two-dimensional neighborhood point set; Identify whether the two-dimensional neighborhood point set is a corner region point set by the distribution characteristics of the points located on both sides of the first feature line; Candidate corner point set acquisition steps: Based on the farthest point pair in the corner region point set, and the farthest point of the second feature line constructed based on the distance from the farthest point pair, identify candidate corner points from the farthest point pair and the farthest point and add them to the candidate corner point set; Final corner point determination steps: For each candidate corner point in the candidate corner point set, the density index and cutoff distance of the candidate corner point are determined based on the third feature line constructed from the farthest point pair and the two non-candidate corner points among the farthest points; the final corner point is determined based on the density index and cutoff distance of all the candidate corner points in the candidate corner point set.

[0007] Secondly, this application provides a corner recognition device based on point clouds, comprising: The boundary point set acquisition module is used to: acquire the surface point cloud of a specific surface of a target object; and acquire a boundary point set based on the surface point cloud to describe the contour of the specific surface. The two-dimensional neighborhood point set acquisition module is used to: extract the three-dimensional neighborhood point set of each boundary point in the boundary point set in the surface point cloud, and obtain the two-dimensional neighborhood point set of the boundary point through dimensionality reduction processing; The corner region point set acquisition module is used to: construct a first feature line based on the farthest point pair in the two-dimensional neighborhood point set; and identify whether the two-dimensional neighborhood point set is a corner region point set by the distribution characteristics of the points located on both sides of the first feature line. The candidate corner point set acquisition module is used to: identify candidate corner points from the farthest point pair and the farthest point of the second feature line constructed based on the farthest point pair in the corner region point set and add them to the candidate corner point set; The final corner point determination module is used to: determine the density index and cutoff distance of the corner region point set corresponding to each candidate corner point in the candidate corner point set, based on the third feature line constructed from the farthest point pair and two non-candidate corner points among the farthest points; and determine the final corner point according to the density index and cutoff distance of all the candidate corner points in the candidate corner point set.

[0008] Thirdly, this application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the corner recognition method based on point clouds provided in the first aspect.

[0009] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the corner recognition method based on point clouds provided in the first aspect.

[0010] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the point cloud-based corner recognition method provided in the first aspect.

[0011] The corner recognition method based on point clouds provided in this application: In the boundary point set acquisition step, the surface point cloud of a specific surface of the target object is acquired, as well as the boundary point set of the surface point cloud. Since the boundary point set can accurately describe the contour of a specific surface of the target object, it lays the foundation for accurate corner point identification. Since corner point identification is based on the boundary point set rather than the entire surface, the amount of data processing is greatly reduced, laying the foundation for rapid corner point identification. In the step of obtaining the two-dimensional neighborhood point set, the three-dimensional neighborhood point set of the extracted boundary points is reduced in dimension to obtain the two-dimensional neighborhood point set of the boundary points. This transforms the complex three-dimensional spatial geometric relationship into a more easily processed two-dimensional planar geometric problem, greatly simplifying the calculation process of subsequent steps. Furthermore, the analysis on the two-dimensional plane can more stably capture the corner features of the contour, is not sensitive to changes in the curvature of the surface point cloud, and avoids the omission and misidentification of corner points. In the corner region point set acquisition step, the first feature line constructed by the farthest point pair is used to quickly locate the potential edge line of the two-dimensional neighborhood point set. By analyzing the distribution characteristics of the points on both sides of the first feature line, it is determined whether the three-dimensional neighborhood point set corresponding to the two-dimensional neighborhood point set is located at the corner of the surface point cloud, that is, whether the two-dimensional neighborhood point set originates from the corner of the surface point cloud, so as to avoid misidentifying points from the straight edge of the surface point cloud as corner points. In the step of obtaining the candidate corner point set, the straight line constructed by the farthest point pair and the farthest point in the corner area point set contains two potential edge straight lines arranged at an angle; the candidate corner points are identified from the farthest point pair and the farthest point to achieve precise locking of the point that best represents the corner point, that is, the candidate corner point. In the final corner point determination step, a third feature line is constructed based on the farthest point pair and the two non-candidate corner points among the farthest points. A density index and cutoff distance are introduced for each candidate corner point to remove false corner points. Since there may be false corner points such as points near the corner point and noise points in the candidate corner point set, the local point distribution density around the false corner point is significantly different from that of the real and clear physical corner point. By using the density index and cutoff distance of all candidate corner points, a final decision is made to eliminate false corner points, avoid misidentification of corner points, and obtain the final corner point, which is the corner point of the contour of a specific surface of the target object.

[0012] Therefore, this application identifies corner point sets originating from the corners of the surface point cloud by utilizing the farthest point pairs within the two-dimensional neighborhood point set of the boundary points. Then, candidate corner points are identified from the farthest point pairs of the corner point set, and the density index and cutoff distance of the candidate corner points are determined. Finally, the final corner point is determined based on the density index and cutoff distance of the candidate corner points, avoiding the occurrence of missed or misidentified corner points. Moreover, the calculation process is simple, achieving accurate and fast corner point identification. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings, wherein the same reference numerals in the following description denote the same parts.

[0015] Figure 1 This is a flowchart illustrating a corner recognition method based on point clouds provided in an embodiment of this application.

[0016] Figure 2A This is a scatter plot of two-dimensional neighborhood points derived from the non-corner, i.e., straight edges of the surface point cloud.

[0017] Figure 2B This is a scatter plot of two-dimensional neighborhood point sets originating from the corners of the surface point cloud.

[0018] Figure 3A First reference point for the corner area point set This is a diagram showing the positional relationship between points and lines when the vertex to be measured is shown.

[0019] Figure 3B Second reference point for the corner area point set This is a diagram showing the positional relationship between points and lines when the vertex to be measured is shown.

[0020] Figure 3C The third reference point for the corner area point set This is a diagram showing the positional relationship between points and lines when the vertex to be measured is shown.

[0021] Figure 4A First reference point for the corner area point set This is a graph showing the distance between points and lines when the vertex to be measured is shown.

[0022] Figure 4B Second reference point for the corner area point set This is a graph showing the distance between points and lines when the vertex to be measured is shown.

[0023] Figure 4C The third reference point for the corner area point set This is a graph showing the distance between points and lines when the vertex to be measured is shown.

[0024] Figure 5A This is a scatter plot of the candidate corner point set obtained based on the boundary point set.

[0025] Figure 5BBased on Figure 5A The decision graph is constructed using density indices and cutoff distances from the candidate corner point set shown.

[0026] Figure 6A This is a scatter plot of corner points identified based on L-shaped surface point cloud.

[0027] Figure 6B This is a scatter plot of corner points identified based on point cloud recognition on a rectangular surface.

[0028] Figure 7 This is another flowchart illustrating a corner recognition method based on point clouds provided in this application embodiment.

[0029] Figure 8 This is a schematic diagram of the corner recognition system based on point cloud provided in the embodiments of this application.

[0030] Figure 9 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the protection scope of this application.

[0032] Figure 1 This is a flowchart illustrating a point cloud-based corner recognition method provided in an embodiment of this application. Please refer to [link / reference]. Figure 1 The corner recognition method based on point cloud provided in this application includes steps 110 to 150, which will be described in detail below.

[0033] Step 110: Obtain the surface point cloud of a specific surface of the target object; based on the surface point cloud, obtain the boundary point set used to describe the contour of the specific surface.

[0034] Step 120: For each boundary point in the boundary point set, extract the three-dimensional neighborhood point set of the boundary point in the surface point cloud, and obtain the two-dimensional neighborhood point set of the boundary point through dimensionality reduction processing.

[0035] Step 130: Construct a first feature line based on the farthest point pair in the two-dimensional neighborhood point set; identify whether the two-dimensional neighborhood point set is a corner region point set by the distribution characteristics of the points located on both sides of the first feature line.

[0036] Step 140: Based on the farthest point pair in the corner region point set and the farthest point of the second feature line constructed based on the distance from the farthest point pair, identify candidate corner points from the farthest point pair and the farthest point and add them to the candidate corner point set.

[0037] Step 150: For each candidate corner point in the candidate corner point set, the density index and cutoff distance of the candidate corner point are determined based on the third feature line constructed from the farthest point pair and the two non-candidate corner points among the farthest points; the final corner point is determined based on the density index and cutoff distance of all the candidate corner points in the candidate corner point set.

[0038] Step 110 can also be called the boundary point set acquisition step; Step 120 can also be called the two-dimensional neighborhood point set acquisition step; Step 130 can also be called the corner region point set acquisition step; Step 140 can also be called the candidate corner point set acquisition step; Step 150 can also be called the final corner point determination step.

[0039] In this embodiment, the surface point cloud of a specific surface of a target object is obtained by sampling the specific surface of the target object using a 3D sensing device such as a LiDAR, depth camera, or 3D scanner. This sampling results in a set of 3D spatial data points used to discretely represent the geometry of the specific surface. These data points are typically represented by coordinates (x, y, z). For descriptive purposes, the surface point cloud of a specific surface of the target object is simply referred to as a surface point cloud.

[0040] The contour of a specific surface of a target object refers to the outer boundary or the boundary of an internal hole of that specific surface of the target object from a specific viewpoint. It reflects the shape of the object from that specific viewpoint. For descriptive purposes, the contour of a specific surface of a target object is simply referred to as the contour.

[0041] A corner point of a contour is a point where the curvature of the contour changes significantly, that is, where the direction of the contour's extension suddenly changes. A corner point is also called a "corner" or "corner point" of the contour. Corner points are key features describing the shape of an object. For descriptive purposes, the corner points of a contour will be simply referred to as corner points in this application.

[0042] In this application, the first feature line, the second feature line, and the third feature line are all straight lines.

[0043] The following description Figure 1 Further optional specific embodiments for each step in the point cloud-based corner recognition process described in the example.

[0044] Step 110: Obtain the surface point cloud of a specific surface of the target object; based on the surface point cloud, obtain the boundary point set used to describe the contour of the specific surface.

[0045] In some embodiments, obtaining a set of boundary points for describing the contour of a specific surface based on the surface point cloud includes: Using α The shape algorithm processes the surface point cloud to obtain the boundary point set.

[0046] For example, α is used The shape algorithm processes the surface point cloud to obtain the boundary point set. .

[0047] Step 120: For each boundary point in the boundary point set, extract the three-dimensional neighborhood point set of the boundary point in the surface point cloud, and obtain the two-dimensional neighborhood point set of the boundary point through dimensionality reduction processing.

[0048] In some embodiments, the step of extracting the three-dimensional neighborhood point set of the boundary point in the surface point cloud includes: Using the three-dimensional coordinates of the boundary point as the center of a sphere, and based on a preset radius or an adaptively determined radius, a three-dimensional neighborhood of the boundary point is constructed in the surface point cloud, wherein the points within the three-dimensional neighborhood of the boundary point constitute the three-dimensional neighborhood point set of the boundary point.

[0049] In the above embodiments, the preset radius can be determined based on the average density of the surface point cloud and the size of a specific surface of the target object. The adaptively determined radius can be dynamically adjusted based on the local point cloud density at the boundary points.

[0050] For example, the bounding sphere algorithm is used to obtain the set of points within the bounding sphere of the boundary points. The bounding sphere of the boundary points is also the 3D neighborhood of the boundary points, and the calculation formula is as follows: ; In the formula, Boundary point The set of points within the bounding sphere, that is, the boundary points. The three-dimensional neighborhood point set; For the set of boundary points; For the set of boundary points The first in One boundary point; Boundary point The radius of the enclosing sphere, Boundary point Points within the enclosing sphere; express It is a point in the surface point cloud; This indicates that the following conditions are met; This indicates the existence of boundary points. Belongs to the boundary point set ; Point To the boundary point The distance is greater than or equal to the radius .

[0051] In some embodiments, the dimensionality reduction process includes: The three-dimensional neighborhood point set of the boundary point is projected onto a reference plane fitted based on the boundary point set to obtain the two-dimensional neighborhood point set of the boundary point.

[0052] Understandably, a corresponding two-dimensional neighborhood point set is obtained based on each boundary point, with a one-to-one correspondence between the boundary point and the two-dimensional neighborhood point set. The distribution of points in the corresponding two-dimensional neighborhood point set differs depending on the location of the boundary point in the surface point cloud. When the boundary point is located at a non-corner point, i.e., a straight edge, in the surface point cloud, the distribution is as follows: Figure 2A The two-dimensional neighborhood point set shown is more likely; when the boundary point is located at the corner of the surface point cloud, the result is as follows: Figure 2B The two-dimensional neighborhood point set shown is more likely.

[0053] For example, a reference plane is obtained by fitting the boundary point set. Project the set of three-dimensional neighborhood points of the boundary points onto the reference plane. This yields the two-dimensional neighborhood point set of the boundary point.

[0054] For example, reference plane for shaft and The two-dimensional neighborhood set of a boundary point on a plane defined by an axis can be represented as: ; in, , Represents the second boundary point within the two-dimensional neighborhood point set. The X-axis and Y-axis coordinates of each point is the number of points in the two-dimensional neighborhood set of the boundary point.

[0055] Step 130: Construct a first feature line based on the farthest point pair in the two-dimensional neighborhood point set; identify whether the two-dimensional neighborhood point set is a corner region point set by the distribution characteristics of the points located on both sides of the first feature line.

[0056] In some embodiments, the step of constructing a first feature line based on the farthest point pair within the two-dimensional neighborhood point set includes: By calculating and comparing the distances of all point pairs in the two-dimensional neighborhood point set, the point pair with the largest distance is determined as the farthest point pair; Connect the farthest point pair to form the first feature line.

[0057] In the above embodiments, the step of calculating the distance between all point pairs in the two-dimensional neighborhood point set includes: Taking each point in the two-dimensional neighborhood point set as the current point, the distance from the current point to each neighboring point in the two-dimensional neighborhood point set is determined using the K-nearest neighbor search algorithm.

[0058] For example, the number of points k in the two-dimensional neighborhood point set is determined, and the distances between the k nearest neighbors of each point in the two-dimensional neighborhood point set are determined using the K-nearest neighbor search algorithm. The calculation formula is as follows: ; In the formula, for Matrix; express The constraints to be satisfied; where i represents the row number, with values ​​from 1 to k, and j represents the column number, with values ​​from 1 to k; The distance between the k-th point and the k-th neighboring point in the two-dimensional neighborhood point set; the k neighboring points of each point include each point itself.

[0059] Compare The size of the distance is used to determine the two points with the largest distance as the two farthest points in the two-dimensional neighborhood point set. and The two farthest points are also known as the farthest point pair; the calculation formula is: ; In the formula, express The two points belong to the two-dimensional neighborhood point set; express The distance between two points; This indicates finding the distance between two points from the two-dimensional neighborhood point set Q. These are the two points corresponding to the maximum value; Based on the two farthest points and Construct a straight line L, which is the first feature line, and calculate it using the following formula: ; In some embodiments, the step of constructing a first feature line based on the farthest point pair within the two-dimensional neighborhood point set includes: Calculate the convex hull of the two-dimensional neighborhood point set; Determine the two farthest vertices from the vertices of the convex hull as the farthest point pair; Connect the farthest point pair to form the first feature line.

[0060] In some embodiments, the step of identifying whether the two-dimensional neighborhood point set is a corner region point set based on the distribution characteristics of points located on both sides of the first feature line includes: Based on the distances from all points in the two-dimensional neighborhood point set to the first feature line, the discreteness of the two-dimensional neighborhood point set is obtained using the discreteness calculation formula. The imbalance of the two-dimensional neighborhood point set is obtained based on the difference in the number of points located on both sides of the first feature line within the two-dimensional neighborhood point set. If the dispersion is greater than or equal to the dispersion threshold, and the imbalance is greater than or equal to the imbalance threshold, then the two-dimensional neighborhood point set is taken as the corner region point set. The dispersion is used to characterize the degree of dispersion of points in the two-dimensional neighborhood point set relative to the first feature line, and the imbalance is used to characterize the degree of uneven distribution of points in the two-dimensional neighborhood point set on both sides of the first feature line.

[0061] For example, the formula for calculating the dispersion can be one of the following: mean absolute deviation, mean absolute deviation, or residual squared. The imbalance of the two-dimensional neighborhood point set can be the ratio of the number of points located on both sides of the first feature line in the two-dimensional neighborhood point set, or it can be the difference in the number of points located on both sides of the first feature line in the two-dimensional neighborhood point set.

[0062] The corner region point set is the two-dimensional neighborhood point set originating from the corner of the surface point cloud, and also the two-dimensional neighborhood point set corresponding to the three-dimensional neighborhood point set located at the corner of the surface point cloud.

[0063] The dispersion threshold is the minimum dispersion of the two-dimensional neighborhood point set when it originates from a corner of the surface point cloud. The imbalance threshold is the minimum imbalance of the two-dimensional neighborhood point set when it originates from a corner of the surface point cloud. Both the dispersion threshold and the imbalance threshold are determined based on the point cloud distance resolution of the surface point cloud. Point cloud distance resolution refers to the density of spatial sampling performed by the 3D sensing device on a specific surface of the target object, that is, the spatial distance between two adjacent sampling points in the point cloud obtained by spatial sampling.

[0064] In the above embodiments, the greater the dispersion, the greater the difference in distance from the points in the two-dimensional neighborhood point set to the first feature line, and the greater the probability that the points in the two-dimensional neighborhood point set are located at the corners of the surface point cloud; the smaller the dispersion, the smaller the difference in distance from the points in the two-dimensional neighborhood point set to the first feature line, and the greater the probability that the two-dimensional neighborhood point set is located at the straight edges of the surface point cloud. The greater the imbalance, the greater the difference in the number of points on both sides of the first feature line within the two-dimensional neighborhood point set, and the greater the probability that the two-dimensional neighborhood point set is located at a corner of the surface point cloud; the smaller the imbalance, the smaller the difference in the number of points on both sides of the first feature line within the two-dimensional neighborhood point set, and the greater the probability that the two-dimensional neighborhood point set is located at a straight edge of the surface point cloud.

[0065] In the above embodiments, by jointly analyzing the dispersion and imbalance of the two-dimensional neighborhood point set, the corner region point set can be identified, which can significantly improve the accuracy of the corner region point set determination, thereby improving the accuracy of subsequent corner point identification. It is understood that if the corner region point set is identified solely by analyzing the dispersion of the two-dimensional neighborhood point set, it cannot rule out the possibility that the two-dimensional neighborhood point set contains a very small number of outlier noise points, which increase the dispersion and thus misidentify it as the corner region point set. Similarly, if the corner region point set is identified solely by analyzing the imbalance of the two-dimensional neighborhood point set, it cannot rule out the possibility that the two-dimensional neighborhood point set originates from slightly curved portions of the straight edges of the surface point cloud, resulting in higher imbalance and thus being misidentified as the corner region point set.

[0066] In some embodiments, the step of obtaining the discreteness of the two-dimensional neighborhood point set based on the distances of all points in the two-dimensional neighborhood point set to the first feature line using a discreteness calculation formula includes: Calculate the distance from each point in the two-dimensional neighborhood point set to the first feature line to obtain the deviation distance of each point; use the standard error calculation formula to calculate the standard error of the deviation distance of all points, and use the standard error as the dispersion of the two-dimensional neighborhood.

[0067] For example, the formula for calculating the distance from each point in the two-dimensional neighborhood point set to the first feature line is as follows: ; In the formula, The third in the three-dimensional neighborhood The distance from the nth point to the first feature line is also the distance of the nth point in the three-dimensional neighborhood. The perpendicular distance from the nth point to the first feature line is also called the nth point in the three-dimensional neighborhood. Deviation distance at each point; This represents the number of points within the three-dimensional neighborhood, which is the total number of points contained in the three-dimensional neighborhood. , These are the coordinates of the two points midway between the farthest point; The third in the three-dimensional neighborhood The coordinates of the points.

[0068] For example, the calculation of the deviation distance of all points The calculation formula is: ; = ; In the formula, The mean square error of the deviation distance between all points in the three-dimensional neighborhood; This is the sum of the squares of the deviation distances of all points.

[0069] In some embodiments, the step of obtaining the imbalance of the two-dimensional neighborhood point set based on the difference in the number of points located on both sides of the first feature line within the two-dimensional neighborhood point set includes: Based on the first and second points in the farthest point pair within the two-dimensional neighborhood point set, a first direction vector is constructed. For each point in the two-dimensional neighborhood point set, a second direction vector is constructed based on the first point and the point itself; Calculate the two-dimensional cross product of the first direction vector and the second direction vector; determine the orientation of the point relative to the first feature line based on the sign of the two-dimensional cross product, where the orientation includes the first side, the second side, or collinearity; count the number of first points located on the first side and the number of second points located on the second side respectively. Based on the first number of points and the second number of points, the imbalance of the two-dimensional neighborhood point set is calculated.

[0070] It can be understood that the first side is to the left of the first feature line, and the second side is to the right of the first feature line; or the first side is to the right of the first feature line, and the second side is to the left of the first feature line. Collinearity means that the point lies on the line of the first feature line.

[0071] In some embodiments, the step of calculating the imbalance of the two-dimensional neighborhood point set based on the first number of points and the second number of points includes: Calculate the ratio of the larger of the first number of points and the second number of points to the smaller of the first number of points and the second number of points, and use the ratio as the imbalance of the two-dimensional neighborhood point set.

[0072] For example, based on the first point in the farthest point pair With the second point The coordinates are used to construct the first direction vector. For each point in the two-dimensional neighborhood Construct the second direction vector The two-dimensional cross product of the first direction vector and the second direction vector is calculated using the following formula: ; For each point The sign of the two-dimensional cross product determines the value of the point. Relative to the first feature line If the location is left, right, or collinear, count the number of points located on the left, right, or collinear respectively. The calculation formula is: ; In the formula, For located at the first feature line The number of dots on the left; For located at the first feature line The number of dots on the right; For located at the first feature line The number of points on it.

[0073] The imbalance of the two-dimensional neighborhood point set is calculated using the following formula: ; In the formula, The number of points on the left. The number of points on the right. The imbalance of the two-dimensional neighborhood point set; To obtain the maximum value, To obtain the minimum value.

[0074] It is understandable that if the two-dimensional neighborhood point set originates from the corner of the surface point cloud, then the points within the two-dimensional neighborhood point set are on the first feature line. The difference in the number of points on the left and right sides is small, indicating a relatively large imbalance in the two-dimensional neighborhood point set. If the two-dimensional neighborhood point set originates from a non-corner area of ​​the surface point cloud, that is, from a straight edge of the surface point cloud, then the points within the two-dimensional neighborhood point set are on the first feature line. The difference in the number of points on the left and right sides is small, and the imbalance of the two-dimensional neighborhood point set is small.

[0075] The formula for determining whether to include the two-dimensional neighborhood point set as the corner region point set is as follows: ; In the formula, The discreteness of the two-dimensional neighborhood point set is... The imbalance of the two-dimensional neighborhood point set; The discreteness threshold, The imbalance threshold; This represents the logical conjunction symbol.

[0076] Step 140: Based on the farthest point pair in the corner region point set and the farthest point of the second feature line constructed based on the distance from the farthest point pair, identify candidate corner points from the farthest point pair and the farthest point and add them to the candidate corner point set.

[0077] In some embodiments, the step of identifying candidate corner points and adding them to the candidate corner point set based on the farthest point pair within the corner region point set and the farthest point of the second feature line constructed based on the distance from the farthest point pair includes: For each set of points in the corner region, determine the farthest pair of points in the set of points in the corner region, including a first reference point and a second reference point; construct a second feature line based on the farthest pair of points, and select the point in the set of points in the corner region that is farthest from the second feature line as a third reference point; The testing steps include: taking the first reference point, the second reference point, and the third reference point as vertices to be tested, constructing a first straight line with the vertex to be tested and the first point of the remaining two reference points, and constructing a second straight line with the second point; calculating the shortest distance from each point in the corner area point set to the first straight line and the second straight line, and calculating the average of the shortest distances of all points; The screening step includes: selecting the vertex to be tested with the smallest average value as a candidate corner point and adding it to the candidate corner point set.

[0078] In the above embodiments, the geometric configuration and distance calculation of the three vertices to be tested are completed through the testing step; and the candidate corner points are selected through the screening step. The selection logic of the candidate corner points is that the set of points in the corner region should be more closely distributed on the two straight lines drawn from the candidate corner points.

[0079] For example, please refer to Figure 3A , Figure 3B and Figure 3C , and For the farthest pair of points in the corner region point set, To base on the farthest point and The constructed second characteristic line, This is the farthest point in the set of points in the corner region, which is also the point where the set of points in the corner region reaches the second feature line. The point with the largest distance, where, As the first reference point, As the second reference point, This serves as the third reference point.

[0080] For example, for , and The following tests were conducted at these three points: Please see Figure 3A ,Will Construct a straight line for the vertex to be measured. and straight lines ; Calculate any point within the point set of the corner region respectively to the straight line The distance to, and the point to the straight line Calculate the distance between the two distances and choose the shortest one. ; Similarly, please refer to Figure 3B ,Will As the vertex to be tested, the points in the corner region are obtained from the straight line. and straight lines shortest distance ; Please see Figure 3C ,Will As the vertex to be tested, the points in the corner region are obtained from the straight line. and straight lines shortest distance The calculation formula is: ; Calculate the shortest distance for all points. average The shortest distance between all points average The shortest distance between all points average The calculation formula is: ; Based on the above calculation formula, the point is obtained. , and For the three point-to-line distance scenarios corresponding to the vertex to be measured, please refer to [link / reference]. Figure 4A , Figure 4B and Figure 4C .

[0081] Understandably, when the vertex to be tested is not a candidate corner point, one of the two constructed lines will inevitably deviate from the boundary point. Please refer to [link / reference needed]. Figure 3A and Figure 3C This results in both the obtained point-to-line distance and the average point-to-line distance being relatively large. Please refer to [link / reference]. Figure 4A and Figure 4C When the vertex to be tested is a candidate corner point, both constructed lines will follow the boundary points. Please refer to [link to relevant documentation]. Figure 3B This results in both the point-to-line distance and the average point-to-line distance being relatively small. Please refer to [link / reference]. Figure 4B Based on this, the vertex to be tested that has the smallest average value of the shortest distance is selected as the candidate corner point. The calculation formula is: ; In the formula: These are candidate corner points.

[0082] Step 150: For each candidate corner point in the candidate corner point set, the density index and cutoff distance of the candidate corner point are determined based on the third feature line constructed from the farthest point pair and the two non-candidate corner points among the farthest points; the final corner point is determined based on the density index and cutoff distance of all the candidate corner points in the candidate corner point set.

[0083] In some embodiments, the step of determining the density index and cutoff distance of the candidate corner point based on the third feature line constructed from the farthest point pair and two non-candidate corner points among the farthest points in the corner region point set corresponding to each candidate corner point in the candidate corner point set includes: The density index calculation steps include: for each candidate corner point, constructing a third feature line based on the farthest point pair in the corner region point set corresponding to the candidate corner point and two non-candidate corner points among the farthest points; determining the distance from the candidate corner point to the third feature line, and using the distance as the density index of the candidate corner point; The cutoff distance calculation step includes: determining the cutoff distance of the candidate corner point based on the density index of each candidate corner point and the distance between the candidate corner point and other candidate corner points in the candidate corner point set other than the candidate corner point itself.

[0084] In the above embodiments, the third feature line constructed based on the farthest point pair and the two non-candidate corner points among the farthest points refers to: for each candidate corner point corresponding to the corner region point set, determining the farthest point pair and the farthest point of the corner region point set, and determining the candidate corner point of the corner region point set, where the candidate corner point is one of the farthest point pair and the farthest point; referring to the two points other than the candidate corner point among the three points of the farthest point pair and the farthest point of the corner region point set as non-candidate corner points, and constructing the third feature line based on these two non-candidate corner points.

[0085] In the above embodiments, each candidate corner point in the candidate corner point set corresponds to a corner region point set, and the candidate corner point originates from the corresponding corner region point set.

[0086] In some embodiments, the step of determining the cutoff distance of a candidate corner point based on the density index of each candidate corner point and the distance between the candidate corner point and other candidate corner points in the candidate corner point set, excluding the candidate corner point itself, includes: The density index of the current candidate corner point is denoted as the first density index; candidate corner points with a density index greater than the first density index are found to obtain a competitive candidate corner point set. Calculate the distance between the current candidate corner point and each of the competing candidate corner points, and take the minimum value among all distances as the cutoff distance of the current candidate corner point.

[0087] In some embodiments, the step of determining the final corner point based on the density index and cutoff distance of all candidate corner points in the candidate corner point set includes: Based on the density index and cutoff distance of all candidate corner points in the candidate corner point set, a decision graph is constructed with the density index as the first coordinate and the cutoff distance as the second coordinate, wherein the first coordinate and the second coordinate are perpendicular to each other; in the decision graph, the candidate corner points that are far away from the first coordinate and the second coordinate are determined as the final corner points.

[0088] In some embodiments, the step of determining the final corner point based on the density index and cutoff distance of all candidate corner points in the candidate corner point set includes: Based on the density index and cutoff distance of all candidate corner points in the candidate corner point set, a decision graph is constructed with the density index as the first coordinate and the cutoff distance as the second coordinate, wherein the first coordinate and the second coordinate are perpendicular to each other; in the decision graph, the candidate corner point that simultaneously satisfies the density index being greater than the first threshold and the cutoff distance being greater than the second threshold is determined as the final corner point.

[0089] Understandably, since points near the candidate corner or noise points may become candidate corners, the accurate corner is obtained by acquiring the density index and cutoff distance of all candidate corners.

[0090] For example, the density index of candidate corner points is calculated. For each candidate corner point, the farthest point pair and the farthest point in the corresponding candidate corner point set are determined. Candidate corner points are then selected from the three points formed by the farthest point pair and the farthest point. The remaining two points are called non-candidate corner points. A third feature line L is constructed based on the two non-candidate corner points. The distance from the candidate corner point to the third feature line L is used as the density index of the candidate corner points. Density index The larger the value, the more likely it is to be a corner point.

[0091] Calculate the cutoff distance of the candidate corner points. Based on the density index... Introducing cutoff distance The concept. All candidate corner points are ranked according to density index. Sort the candidate corner points from largest to smallest. Then, process the candidate corner points sequentially, starting with the one with the highest density index.

[0092] Calculate the cutoff distance of the candidate corner points. For each candidate point in the candidate point set, perform the following steps: for the current candidate corner point... Obtain the density index of the current candidate corner point i. Determine the density index is greater than All candidate corner points At the same time, all candidate corner points are obtained. density index When the density index is greater than All candidate corner points In the middle, calculate the current candidate corner point To each candidate corner point distance , all distances The minimum value in the range is used as the current candidate corner point. Cutoff distance The calculation formula is: ; In the formula: For the current candidate corner point The cutoff distance; For the current candidate corner point To candidate corner The distance; This indicates that the search density index is greater than that of the current candidate corner. Candidate corners of density index ; For the current candidate corner point Density index, Candidate corner points Density index.

[0093] Please see Figure 5A and Figure 5B Based on the density index of all candidate corner points in the candidate corner point set and cutoff distance Construct a decision graph. See also Figure 6A and Figure 6B Select cutoff distance and density index The candidate corner point with the larger value is the final corner point.

[0094] Please see Figure 7 In some embodiments, the corner recognition method based on point clouds includes: The boundary point set acquisition steps include: acquiring the surface point cloud of a specific surface of the target object, using α... The shape algorithm processes the surface point cloud to obtain a set of boundary points used to describe the contour of a specific surface. The steps for obtaining the two-dimensional neighborhood point set include: obtaining the three-dimensional neighborhood point set of the boundary points using the bounding sphere algorithm; and fitting the boundary point set to obtain the reference plane. Project the set of three-dimensional neighborhood points of the boundary points onto the reference plane. This yields the two-dimensional neighborhood point set of the boundary point; The corner region point set acquisition step includes: taking each point in the two-dimensional neighborhood point set as the current point, using the K-nearest neighbor search algorithm to determine the distance from the current point to each neighboring point in the two-dimensional neighborhood point set, and then determining the farthest point pair in the two-dimensional neighborhood point set; for the two-dimensional neighborhood point set, constructing a first feature line based on the farthest point pair, and identifying whether the two-dimensional neighborhood point set is a corner region point set based on the distance from the points in the two-dimensional neighborhood point set to the first feature line and the difference in the number of points on both sides of the first feature line; The candidate corner point set acquisition step includes: based on the farthest point pair in the corner region point set, and the farthest point of the second feature line constructed based on the distance from the farthest point pair, each point in the farthest point pair and the farthest point is respectively taken as a vertex to be tested; based on the vertex to be tested, a first straight line and a second straight line are constructed; the average of the shortest distances from all points in the corner region point set to the first straight line and the second straight line is calculated; the vertex to be tested with the smallest average value is selected as a candidate corner point and added to the candidate corner point set; The final corner point determination step includes: for each candidate corner point in the candidate corner point set corresponding to the corner region point set, based on the third feature line constructed from the farthest point pair and two non-candidate corner points among the farthest points, determining the density index and cutoff distance of the candidate corner point; based on the density index and cutoff distance of all the candidate corner points in the candidate corner point set, constructing a decision map with the density index as the first coordinate and the cutoff distance as the second coordinate, wherein the first coordinate and the second coordinate are perpendicular to each other; in the decision map, determining the final corner point as the candidate corner point that is far away from the first coordinate and the second coordinate.

[0095] Figure 8 This is a schematic diagram of the corner recognition system based on point clouds provided in the embodiments of this application. Please refer to... Figure 8 The point cloud-based corner recognition system may include a boundary point set acquisition module 801, a two-dimensional neighborhood point set acquisition module 802, a corner region point set acquisition module 803, a candidate corner point set acquisition module 804, and a final corner point determination module 805. Wherein: Boundary point set acquisition module 801 is used to: acquire the surface point cloud of a specific surface of the target object; and acquire a boundary point set to describe the contour of the specific surface based on the surface point cloud.

[0096] The two-dimensional neighborhood point set acquisition module 802 is used to: extract the three-dimensional neighborhood point set of each boundary point in the boundary point set in the surface point cloud, and obtain the two-dimensional neighborhood point set of the boundary point through dimensionality reduction processing.

[0097] The corner area point set acquisition module 803 is used to: construct a first feature line based on the farthest point pair in the two-dimensional neighborhood point set; and identify whether the two-dimensional neighborhood point set is a corner area point set by the distribution characteristics of the points located on both sides of the first feature line.

[0098] The candidate corner point set acquisition module 804 is used to: identify candidate corner points from the farthest point pair and the farthest point of the second feature line constructed based on the farthest point pair in the corner region point set and add them to the candidate corner point set.

[0099] The final corner point determination module 805 is used to: determine the density index and cutoff distance of the corner region point set corresponding to each candidate corner point in the candidate corner point set based on the third feature line constructed from the farthest point pair and the two non-candidate corner points in the farthest point; and determine the final corner point based on the density index and cutoff distance of all the candidate corner points in the candidate corner point set.

[0100] In practical applications, the above system can be a terminal device or a chip applied to a terminal device. In this application, the system can implement the functions of multiple units through software, hardware, or a combination of software and hardware, enabling the system to execute the corner recognition method steps based on point clouds as provided in any of the above embodiments. Furthermore, the technical effects of each technical solution of this system can be referenced to the technical effects of the corresponding technical solutions in the corner recognition method based on point clouds, and will not be elaborated upon further in this application.

[0101] Figure 9 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0102] Based on the hardware implementation of each unit in the above-described device, embodiments of this application also provide an electronic device, such as... Figure 9 As shown, the electronic device 900 includes a memory 910 and a processor 920. The memory 910 stores a computer program, and the processor 920 executes the computer program to implement the steps of the corner recognition method based on point cloud provided in any of the above embodiments.

[0103] Of course, in practical applications, such as Figure 9 As shown, the various components in the electronic device 900 are coupled together via a bus system 930. It can be understood that the bus system 930 is used to enable communication between these components. In addition to a data bus, the bus system 930 also includes a power bus, a control bus, and a status signal bus. However, for clarity, all buses are labeled as bus system 930 in the figure.

[0104] In practical applications, the aforementioned processor can be at least one of the following: Application-Specific Integrated Circuit (ASIC), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field-Programmable Gate Array (FPGA), controller, microcontroller, and microprocessor. It is understood that, for different devices, the electronic devices used to implement the functions of the aforementioned processor can also be other types, and the embodiments of this application do not specifically limit this.

[0105] The aforementioned memory can be volatile memory, such as random-access memory (RAM); or non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); or a combination of the above types of memory, and provides instructions and data to the processor.

[0106] The electronic devices described in the embodiments of this application can be terminal devices or chips applied to terminal devices. The terminal devices described in the embodiments of this application can include terminal devices such as mobile phones, tablet computers, laptops, handheld computers, personal digital assistants (PDAs), portable media players (PMPs), wearable devices, cameras, etc.

[0107] In an exemplary embodiment, this application also provides a computer-readable storage medium, such as a memory including a computer program, which can be executed by a processor of an electronic device to perform the steps of the aforementioned method.

[0108] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any one of the embodiments of this application.

[0109] Optionally, the computer program product can be applied to the electronic device in the embodiments of this application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the electronic device in the various methods of the embodiments of this application. For the sake of brevity, they will not be described in detail here.

[0110] This application also provides a computer program.

[0111] Optionally, the computer program can be applied to the electronic device in the embodiments of this application. When the computer program is run on a computer, it causes the computer to execute the corresponding processes implemented by the electronic device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0112] It should be understood that in the embodiments of this application, data such as user information are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0113] It should be understood that the terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items. The expressions “having,” “may have,” “comprising,” and “including,” or “may include” and “may contain” used herein may be used to indicate the presence of a corresponding feature (e.g., an element such as a number, function, operation, or component), but do not exclude the presence of additional features.

[0114] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another, and are not necessarily used to describe a specific order or sequence. For example, without departing from the scope of this invention, first information may also be referred to as second information, and similarly, second information may also be referred to as first information.

[0115] The technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0116] In the several embodiments provided in this application, it should be understood that the disclosed methods, systems, and devices can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0117] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0118] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0119] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A corner recognition method based on point clouds, characterized in that, include: Obtain the surface point cloud of a specific surface of a target object; Based on the surface point cloud, a set of boundary points is obtained to describe the contour of a specific surface; For each boundary point in the boundary point set, extract the three-dimensional neighborhood point set of the boundary point in the surface point cloud, and obtain the two-dimensional neighborhood point set of the boundary point through dimensionality reduction processing; A first feature line is constructed based on the farthest point pair within the two-dimensional neighborhood point set; the distribution characteristics of points located on both sides of the first feature line are used to identify whether the two-dimensional neighborhood point set is a corner region point set. Based on the farthest point pair within the corner region point set, and the farthest point of the second feature line constructed based on the distance from the farthest point pair, candidate corner points are identified from the farthest point pair and the farthest point and added to the candidate corner point set; For each candidate corner point in the candidate corner point set, the density index and cutoff distance of the candidate corner point are determined based on the third feature line constructed from the farthest point pair and the two non-candidate corner points among the farthest points; the final corner point is determined based on the density index and cutoff distance of all the candidate corner points in the candidate corner point set.

2. The corner recognition method based on point clouds according to claim 1, characterized in that, The dimensionality reduction process includes the following steps: The three-dimensional neighborhood point set of the boundary point is projected onto a reference plane fitted based on the boundary point set to obtain the two-dimensional neighborhood point set of the boundary point.

3. The corner recognition method based on point clouds according to claim 1, characterized in that, The step of identifying whether the two-dimensional neighborhood point set is a corner region point set based on the distribution characteristics of points located on both sides of the first feature line includes: Based on the distances from all points in the two-dimensional neighborhood point set to the first feature line, the discreteness of the two-dimensional neighborhood point set is obtained using the discreteness calculation formula. The imbalance of the two-dimensional neighborhood point set is obtained based on the difference in the number of points located on both sides of the first feature line within the two-dimensional neighborhood point set. If the dispersion is greater than or equal to the dispersion threshold, and the imbalance is greater than or equal to the imbalance threshold, then the two-dimensional neighborhood point set is taken as the corner region point set. The dispersion is used to characterize the degree of dispersion of points in the two-dimensional neighborhood point set relative to the first feature line, and the imbalance is used to characterize the degree of uneven distribution of points in the two-dimensional neighborhood point set on both sides of the first feature line.

4. The corner recognition method based on point clouds according to claim 3, characterized in that, The step of obtaining the imbalance degree of the two-dimensional neighborhood point set based on the difference in the number of points located on both sides of the first feature line within the two-dimensional neighborhood point set includes: Based on the first point and the second point in the farthest point pair within the two-dimensional neighborhood point set, a first direction vector is constructed; for each point in the two-dimensional neighborhood point set, a second direction vector is constructed based on the first point and the point itself. Calculate the two-dimensional cross product of the first direction vector and the second direction vector; determine the orientation of the point relative to the first feature line based on the sign of the two-dimensional cross product, the orientation including the first side, the second side, or collinearity; count the number of first points located on the first side and the number of second points located on the second side respectively; calculate the imbalance of the two-dimensional neighborhood point set based on the number of first points and the number of second points.

5. The corner recognition method based on point clouds according to claim 1, characterized in that, The step of identifying candidate corner points and adding them to the candidate corner point set based on the farthest point pair within the corner region point set and the farthest point of the second feature line constructed based on the distance from the farthest point pair includes: For each set of points in the corner region, determine the farthest pair of points in the set of points in the corner region, including a first reference point and a second reference point; construct a second feature line based on the farthest pair of points, and select the point in the set of points in the corner region that is farthest from the second feature line as a third reference point; The testing steps include: taking the first reference point, the second reference point, and the third reference point as vertices to be tested, constructing a first straight line with the vertex to be tested and the first point of the remaining two reference points, and constructing a second straight line with the second point; calculating the shortest distance from each point in the corner area point set to the first straight line and the second straight line, and calculating the average of the shortest distances of all points; The screening step includes: selecting the vertex to be tested with the smallest average value as a candidate corner point and adding it to the candidate corner point set.

6. The corner recognition method based on point clouds according to claim 1, characterized in that, The step of determining the density index and cutoff distance of each candidate corner point in the candidate corner point set, based on the third feature line constructed from the farthest point pair and two non-candidate corner points among the farthest points, includes: The density index calculation steps include: for each candidate corner point, constructing a third feature line based on the farthest point pair in the corner region point set corresponding to the candidate corner point and two non-candidate corner points among the farthest points; determining the distance from the candidate corner point to the third feature line, and using the distance as the density index of the candidate corner point; The cutoff distance calculation step includes: determining the cutoff distance of the candidate corner point based on the density index of each candidate corner point and the distance between the candidate corner point and other candidate corner points in the candidate corner point set other than the candidate corner point itself.

7. The corner recognition method based on point clouds according to claim 6, characterized in that, The step of determining the cutoff distance of a candidate corner point based on the density index of each candidate corner point and the distance between the candidate corner point and other candidate corner points in the candidate corner point set, excluding the candidate corner point itself, includes: The density index of the current candidate corner point is denoted as the first density index; candidate corner points with a density index greater than the first density index are found to obtain a competitive candidate corner point set. Calculate the distance between the current candidate corner point and each of the competing candidate corner points, and take the minimum value among all distances as the cutoff distance of the current candidate corner point.

8. The corner recognition method based on point clouds according to claim 1, characterized in that, The step of determining the final corner point based on the density index and cutoff distance of all candidate corner points in the candidate corner point set includes: Based on the density index and cutoff distance of all candidate corner points in the candidate corner point set, a decision graph is constructed with the density index as the first coordinate and the cutoff distance as the second coordinate, wherein the first coordinate and the second coordinate are perpendicular to each other; in the decision graph, the candidate corner points that are far away from the first coordinate and the second coordinate are determined as the final corner points.

9. A corner recognition system based on point clouds, characterized in that, include: The boundary point set acquisition module is used to: acquire the surface point cloud of a specific surface of a target object; Based on the surface point cloud, a set of boundary points is obtained to describe the contour of a specific surface; The two-dimensional neighborhood point set acquisition module is used to: extract the three-dimensional neighborhood point set of each boundary point in the boundary point set in the surface point cloud, and obtain the two-dimensional neighborhood point set of the boundary point through dimensionality reduction processing; The corner region point set acquisition module is used to: construct a first feature line based on the farthest point pair in the two-dimensional neighborhood point set; and identify whether the two-dimensional neighborhood point set is a corner region point set by the distribution characteristics of the points located on both sides of the first feature line. The candidate corner point set acquisition module is used to: identify candidate corner points from the farthest point pair and the farthest point of the second feature line constructed based on the farthest point pair in the corner region point set and add them to the candidate corner point set; The final corner point determination module is used to: determine the density index and cutoff distance of the corner region point set corresponding to each candidate corner point in the candidate corner point set, based on the third feature line constructed from the farthest point pair and two non-candidate corner points among the farthest points; and determine the final corner point according to the density index and cutoff distance of all the candidate corner points in the candidate corner point set.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the corner recognition method based on point clouds provided in any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the corner recognition method based on point clouds provided in any one of claims 1 to 8.

Images