An adaptive three-dimensional circular hole feature recognition method based on alpha-shape

Abstract

The application provides an adaptive three-dimensional circular hole feature recognition method based on an alpha-shape, wherein the method calculates a contour extraction control factor alpha according to the density of a point set to achieve the purpose of adaptive contour extraction of the point set; a conditional filtering method is used to filter out boundary points and discrete noise points in the contour points which do not belong to the circular hole; and a Ransac method is used to fit the contour points to obtain three-dimensional circular hole center, radius and normal feature information. The method can be applied to a complex circular hole feature recognition scene, has strong robustness and has high positioning accuracy.

Description

Technical Field

[0001] This invention belongs to the field of three-dimensional visual recognition and relates to an adaptive three-dimensional circular hole feature recognition method based on α-shape. Background Technology

[0002] 3D circular hole feature recognition is a technology based on 3D point set data, designed to accurately and efficiently identify key features such as the location, size, and shape of circular holes in complex 3D scenes. Circular holes are common component features in many engineering applications, and their geometry and location information significantly impact product quality and performance. Therefore, accurate and rapid feature recognition and measurement of 3D circular holes is a crucial step in ensuring product quality and production efficiency.

[0003] 3D circular hole feature recognition is widely used in industrial manufacturing, robot vision, and automation. In industrial manufacturing, it can be applied to the production and assembly processes of various workpieces. For example, in automobile manufacturing, automotive parts typically have many circular holes, such as screw holes and pin holes. 3D circular hole feature recognition technology can quickly and accurately detect the position, diameter, and depth of these holes, providing accurate positioning information for automated assembly. In robot vision, 3D circular hole feature recognition can be applied to robot positioning and grasping tasks. By recognizing the circular hole features on a target object, the robot can accurately determine the position and orientation of the grasping point, achieving efficient automated assembly and processing operations. For example, on an industrial robot assembly line, 3D circular hole feature recognition technology allows the robot to quickly and accurately locate screw holes on parts, enabling automated screw tightening operations.

[0004] In conclusion, research on 3D circular hole feature recognition algorithms has significant theoretical and practical implications for the development of machine vision and image processing. This research can not only improve the accuracy and efficiency of circular hole recognition, but also provide broader ideas and methods for the research and application of automated vision systems.

[0005] Currently, methods for recognizing 3D circular holes include those based on geometric features and those based on machine learning. Geometric feature-based methods require significant manual intervention in setting parameters, resulting in poor flexibility; machine learning-based methods require substantial computational resources and time, and the model training process is cumbersome, leading to poor real-time performance. Therefore, a new method for recognizing 3D circular holes is urgently needed to address these issues. Summary of the Invention

[0006] The purpose of this invention is to solve the problem of fast and accurate adaptive feature recognition of three-dimensional circular holes, and to propose an adaptive three-dimensional circular hole feature recognition method based on α-shape.

[0007] This invention is achieved using the following technical solution:

[0008] An adaptive 3D circular hole feature recognition method based on α-shape, the steps of which are as follows:

[0009] (1) Construct the topological relationship of the point set. Obtain the point set P{p} in the XYZ coordinate system. i (x i ,y i ,z i )|i=1,2,...,n}, construct a kd-tree based on the coordinate relationships of the points in the point set to speed up the neighborhood search of the point set;

[0010] (2) Adaptively calculate the control factor α. Let k = n / 1000, for each point p in the point set P... i The k nearest neighbors of a node are searched using a kd-tree, forming a set P. i {p ij (x ij ,y ij ,z ij The control factor α is calculated using equations (1) and (2) for each j = 1, 2, ..., k.

[0011]

[0012]

[0013] (3) Point set projection. A plane is obtained by least-squares plane fitting based on the points in the point set P. The normal vector of this plane is... Project each point in P onto the plane, and rotate the plane until it is parallel to the XOY plane in the XYZ coordinate system to obtain the projection point set P'{p' i (x′ i ,y′ i ,z′ i |i=1,2,...,n};

[0014] (4) Delaynay triangulation. The projected point set P' is triangulated using Delaynay to obtain the mesh of connections between each point in the point set P;

[0015] Step 1: Query point p' in the projection point set. i The initial side length AB is obtained by connecting the nearest point A and the second nearest point B on the Euclidean distance.

[0016] Step 2, construct the initial triangle with the initial side length AB as the side to be expanded: set the search radius r, draw a circle with the midpoint of the initial side length AB as the center and r as the radius, and construct the initial triangle ΔABC by taking any projection point C in the search circle area as the third point.

[0017] Step 3: Search for other projection points outside ΔABC, and the projection points and the third point outside the side to be expanded of ΔABC are located on both sides of the side to be expanded; using the side lengths AB, AC and BC of the initial triangle ΔABC as the side to be expanded, repeat Step 2 to obtain the expanded triangle;

[0018] Step 4: In the point set P, find the triangle formed by the three points corresponding to the three vertices of the initial triangle ΔABC and the triangle formed by the three points corresponding to the three vertices of the extended triangle, and calculate the angle between the normal vectors of the two (specifically, the triangle corresponding to the initial triangle and the triangle corresponding to the extended triangle), and select the triangle with the largest angle between the normal vectors as the extended triangle.

[0019] Step 5: If multiple triangles satisfy the conditions in Step 4, compare their smallest interior angles and select the triangle with the largest smallest interior angle as the extended triangle.

[0020] Step 6: Using the three sides of the extended triangle as the sides to be extended, continue to extend according to Step 3-Step 5 until the connection relationship of all points is determined.

[0021] (5) Extract boundary points using the α-shape algorithm. Traverse each edge in the Mesh obtained after Delaunay meshing. If the length of the edge is less than 2α, draw a circle with radius α around the edge. If the circle contains other points, delete the edge from the Mesh. The remaining edges in the Mesh are the boundary of the point set, and the points on the boundary are the boundary points.

[0022] (6) Point set conditional filtering. Calculate the minimum and maximum values ​​of the x and y coordinates of the boundary points: Min_x, Max_x, Min_y, Max_y. Set a separation interval: only retain points whose x coordinate values ​​are in [Min_x+interval, Max_x-interval] and whose y coordinate values ​​are in [Min_y+interval, Max_y-interval], thereby removing discrete noise points and leaving only the boundary points of the circular holes.

[0023] (7) Ransac fitting. Perform Ransac fitting on the boundary points of the remaining point set after filtering in step (6) to output the center coordinates, radius and normal parameters of the circular hole.

[0024] The beneficial effects of this invention are as follows:

[0025] This invention addresses the problem of adaptive feature recognition of three-dimensional circular holes by proposing an α-shape-based adaptive three-dimensional circular hole feature recognition method. A contour extraction control factor α is calculated based on the density of the point set to achieve adaptive contour extraction. Conditional filtering is used to filter out boundary points that do not belong to the circular hole and discrete noise points from the contour points. The Ransac method is then used to fit the contour points to obtain the center, radius, and normal feature information of the three-dimensional circular hole. This method is applicable to complex circular hole feature recognition scenarios, exhibits strong robustness, and has high positioning accuracy. Attached Figure Description

[0026] Figure 1 Flowchart of the steps for calculating the control factor α for contour extraction;

[0027] Figure 2 Flowchart of steps for extracting the outline of a circular hole for α-shape;

[0028] Figure 3 Comparison of contour extraction effects for different α-circular hole point sets;

[0029] Figure 4 This is a flowchart showing the actual 3D circular hole contour extraction process.

[0030] Figure 5 This is a diagram showing the steps of Delaynay triangulation. Detailed Implementation

[0031] This invention provides an adaptive three-dimensional circular hole feature recognition method based on α-shape, the specific steps of which are as follows:

[0032] (1) Construct the topological relationship of the point set. Obtain the point set P{p} in the XYZ coordinate system. i (x i ,y i ,z i For each i = 1, 2, ..., n, construct a kd-tree based on the coordinate relationships of the points in the point set to speed up the neighborhood search of the point set.

[0033] (2) Adaptive calculation of control factor α, the specific process is as follows: Figure 1 As shown. Let k = n / 1000, for each point p in the point set P. i The k nearest neighbors of a node are searched using a kd-tree, forming a set P. i {p ij (x ij ,y ij ,z ijThe control factor α is calculated using equations (1) and (2) for each j = 1, 2, ..., k.

[0034]

[0035]

[0036] (3) Point set projection. A plane is obtained by least-squares plane fitting based on the points in the point set P. The normal vector of this plane is... Project each point in P onto the plane, and rotate the plane until it is parallel to the XOY plane in the XYZ coordinate system to obtain the projection point set P'{p' i (x′ i ,y′ i ,z′ i )|i=1,2,...,n}.

[0037] (4) Delaynay triangulation. The projected point set P' is triangulated using Delaynay to obtain the mesh of connections between each point in the point set P;

[0038] Step 1: Query point p' in the projection point set. i The initial side length AB is obtained by connecting the nearest point A and the second nearest point B on the Euclidean distance.

[0039] Step 2, construct the initial triangle with the initial side length AB as the side to be expanded: set the search radius r, draw a circle with the midpoint of the initial side length AB as the center and r as the radius, and construct the initial triangle ΔABC by taking any projection point C in the search circle area as the third point.

[0040] Step 3: Search for other projection points outside ΔABC, and the projection points and the third point outside the side to be expanded of ΔABC are located on both sides of the side to be expanded; using the side lengths AB, AC and BC of the initial triangle ΔABC as the side to be expanded, repeat Step 2 to obtain the expanded triangle;

[0041] Step 4: In the point set P, find the triangle formed by the three points corresponding to the three vertices of the initial triangle ΔABC and the triangle formed by the three points corresponding to the three vertices of the extended triangle, and calculate the angle between the normal vectors of the two (specifically, the triangle corresponding to the initial triangle and the triangle corresponding to the extended triangle), and select the triangle with the largest angle between the normal vectors as the extended triangle.

[0042] Step 5: If multiple triangles satisfy the conditions in Step 4, compare their smallest interior angles and select the triangle with the largest smallest interior angle as the extended triangle.

[0043] Step 6: Using the three sides of the extended triangle as the sides to be extended, continue to extend according to Step 3-Step 5 until the connection relationship of all points is determined.

[0044] (5) Extract boundary points using the α-shape algorithm. Traverse each edge in the Mesh obtained after Delaunay meshing. If the length of the edge is less than 2α, then draw a circle with radius α around the edge. If the circle contains other points, then delete the edge from the Mesh. Finally, the remaining edges in the Mesh are the boundary of the point set, and the points on the boundary are the boundary points. Figure 2 The flowchart is for steps (4)-(5).

[0045] (6) Point set conditional filtering. Statistically calculate the minimum and maximum x and y coordinates (Min_x, Max_x, Min_y, Max_y) of the boundary points. Set a separation interval (interval), retaining only points where the x-coordinate value falls within [Min_x+interval, Max_x-interval] and the y-coordinate value falls within [Min_y+interval, Max_y-interval]. This removes discrete noise points, leaving only the boundary points of the circular hole. This allows for the extraction of the circular hole's contour (e.g., ...). Figure 4 ).

[0046] (7) Ransac fitting. Perform Ransac fitting on the boundary points of the remaining point set after filtering in step (6) to output the center coordinates, radius and normal parameters of the circular hole.

[0047] To verify the rationality and effectiveness of the adaptive α calculation, a point set was randomly generated by sampling a plane with a square hole of 2*2mm side length and a circular hole radius of 1mm. Figure 3 (a) , manually set the value of α to 0.005, 0.01, 0.1, 0.45 ( Figure 3 (b)-(e)) Boundary extraction is performed on the set of circular hole points, and the adaptive α-shape method is used simultaneously. Figure 3 (f) Automatically calculate α to extract the boundary. After comparing the results, the method proposed in this invention can effectively and adaptively extract the contour points of the circular hole.

[0048] Figure 4 This is a flowchart illustrating the extraction process for the actual circular hole contour. First, a least-squares plane fitting is performed based on the point set, and then the point set is projected onto the fitted plane to obtain the projected point set. Figure 4 (a)), and then the adaptive α-shape contour extraction method proposed in this invention is used to extract the point set contour ( Figure 4 (b) Finally, conditional filtering is used to filter out the outer frame that does not belong to the boundary of the circular hole. Figure 4(c)) to obtain the outline points of the circular hole.

[0049] Figure 5 This is a flowchart of the Delaynay triangulation process. The Delaynay triangulation method establishes the connections between points in the projection point set, providing a prerequisite for the α-shape algorithm to extract the contour of the projection point set. Figure 5 (d) shows the angle between the normal vectors of the triangle corresponding to the initial triangle and the triangle corresponding to the extended triangle.

[0050] Table 1 shows the actual 3D circular hole feature recognition results. The RANSAC method was used to... Figure 4 The obtained circular hole contour points are used to fit the circular hole, resulting in the center coordinates and radius of the hole. The range of each coordinate component of the center coordinate and the radius does not exceed 3μm, indicating that the algorithm is stable, accurate, and reliable.

[0051] Table 1. Results of actual three-dimensional circular hole feature recognition

[0052]

Claims

1. An adaptive three-dimensional circular hole feature recognition method based on α-shape, characterized in that, The steps are as follows: (1) Constructing the topological relationship of the point set: Obtain the point set P{p} in the XYZ coordinate system. i (x i ,y i ,z i Construct a kd-tree based on the coordinate relationships of the points in the point set (i = 1, 2, ..., n). (2) Adaptive calculation of control factor α: Let k = n / 1000, for each point p in the point set P, i The k nearest neighbors of a node are searched using a kd-tree, forming a set P. i {p ij (x ij ,y ij ,z ij The control factor α is calculated using equations (1) and (2) for each j = 1, 2, ..., k. (3) Point set projection: A plane is obtained by least-squares plane fitting based on the points in the point set P, and the normal vector of the plane is... Project each point in P onto the plane, and rotate the plane until it is parallel to the XOY plane in the XYZ coordinate system to obtain the projection point set P'{p' i (x i ',y i ',z i ')|i=1,2,...,n}; (4) Delaynay triangulation: Delaynay triangulation is performed on the projected point set P' to obtain the connection relationship Mesh between each point in the point set P; (5) Extract boundary points using the α-shape algorithm: Traverse each edge in the Mesh obtained after Delaunay meshing. If the length of the edge is less than 2α, then draw a circle with radius α using the edge. If the circle contains other points, then delete the edge from the Mesh. Finally, the remaining edges in the Mesh are the boundary of the point set, and the points on the boundary are the boundary points. (6) Point set conditional filtering: Calculate the minimum and maximum values ​​of x and y coordinates Min_x, Max_x, Min_y, Max_y among the boundary points, set a separation interval interval, and only keep the points whose x coordinate values ​​are in [Min_x+interval, Max_x-interval] and whose y coordinate values ​​are in [Min_y+interval, Max_y-interval]. (7) Ransac fitting: Perform Ransac fitting on the boundary points of the remaining point set after filtering in step (6) to output the center coordinates, radius and normal parameters of the circular hole.

2. The adaptive three-dimensional circular hole feature recognition method based on α-shape according to claim 1, characterized in that, The Delaynay triangulation method in step (4) is as follows: Step 1: Query point p' in the projection point set. i The initial side length AB is obtained by connecting the nearest point A and the second nearest point B on the Euclidean distance. Step 2, construct the initial triangle with the initial side length AB as the side to be expanded: set the search radius r, draw a circle with the midpoint of the initial side length AB as the center and r as the radius, and construct the initial triangle ΔABC by taking any projection point C in the search circle area as the third point. Step 3: Search for other projection points outside ΔABC, and the projection points and the third point outside the side to be expanded of ΔABC are located on both sides of the side to be expanded; using the side lengths AB, AC and BC of the initial triangle ΔABC as the side to be expanded, repeat Step 2 to obtain the expanded triangle; Step 4: In the point set P, find the triangle formed by the three points corresponding to the three vertices of the initial triangle ΔABC and the triangle formed by the three points corresponding to the three vertices of the extended triangle, and calculate the angle between the normal vectors of the two triangles. Select the triangle with the largest angle between the normal vectors as the extended triangle. Step 5: If multiple triangles satisfy the conditions in Step 4, compare their smallest interior angles and select the triangle with the largest smallest interior angle as the extended triangle. Step 6: Using the three sides of the extended triangle as the sides to be extended, continue to extend according to Step 3-Step 5 until the connection relationship of all points is determined.

Images