A method and system for identifying in-out boundary points in a point cloud contour

Abstract

This invention belongs to the field of point cloud model technology and discloses a method and system for identifying inscribed and circumscribed boundary points of point cloud contours. The invention performs planar segmentation on the original point cloud to obtain multiple independent planar point clouds. For each planar point cloud, its projection plane is obtained through planar fitting. The boundary points of the planar point cloud are extracted and projected onto the corresponding projection plane. The projected 3D boundary point set is transformed to the xoy plane to obtain a 2D boundary point set. An envelope rectangle is constructed for the 2D boundary point set using the minimum bounding rectangle algorithm. The 2D boundary point set is then scanned at intervals, and the distance from each 2D boundary point within the scan range to its corresponding rectangle edge is calculated. The 2D boundary point with the smallest distance is identified as an inscribed boundary point, and the remaining 2D boundary points are identified as inscribed boundary points or redundant points. This invention can effectively distinguish between inscribed and circumscribed boundary points.

Description

Technical Field

[0001] This invention belongs to the field of point cloud model technology, and more specifically, relates to a method and system for identifying the inner and outer boundary points of a point cloud contour. Background Technology

[0002] With the rapid development of 3D scanning technologies (such as LiDAR, structured light, and photogrammetry), point clouds, as the most direct 3D digital representation of the physical world, have been widely applied in many fields such as reverse engineering, industrial inspection, autonomous driving, smart cities, and cultural heritage protection. Point cloud data consists of massive amounts of discrete 3D spatial points and does not contain any explicit topological connections. Therefore, accurately and efficiently extracting geometric features from disordered point clouds is one of the core problems in 3D computer vision and graphics. Among the many geometric features, the contour boundary points (or edge points) of the point cloud are key information representing the shape contour, structural features, and topological relationships of an object. Accurate extraction of boundary points is crucial for high-level tasks such as point cloud segmentation, registration, simplification, 3D reconstruction, and feature-based recognition and detection. Generally, boundary points can be divided into external boundary points and internal boundary points. External boundary points are mainly points located at the edges of the external contour of the object described by the point cloud, while internal boundary points refer to edge points formed inside the object's surface (such as the edges of holes or cracks), revealing the detailed structure and intrinsic features of the object's surface.

[0003] Currently, there is a lot of research on point cloud boundary point extraction technology. The mainstream methods mainly include: (1) Methods based on normal vector or curvature change. The principle of this method is intuitive, but it is extremely sensitive to point cloud noise and will generate a lot of noise pseudo-boundaries. More importantly, this type of method usually cannot effectively distinguish whether the extracted boundary points are external or internal boundaries, but mixes them together, which greatly limits the accuracy of subsequent applications. (2) Methods based on α-shape or convex hull. This method can only extract the outermost convex hull boundary and cannot handle concave boundaries and internal boundaries. It also does not provide category information for internal and external boundaries. (3) Methods based on projection or clustering. This type of method will inevitably introduce the loss and distortion of three-dimensional information, resulting in inaccurate boundary extraction. It also faces the problem of not being able to automatically distinguish between internal and external boundaries. (4) Methods based on deep learning. This method performs well on some datasets, but it relies on a large amount of high-quality labeled data for training. The generalization ability of the model is limited. For point clouds with unseen object types or large differences in scanning conditions, the performance may drop sharply. Moreover, its output does not distinguish between internal and external boundaries. In summary, the main drawback of existing technologies is their inability to distinguish between internal and external boundary points, resulting in the output of a mixed external contour and internal structural boundary, which fails to meet the application requirements for detailed structural analysis. Summary of the Invention

[0004] This invention provides a method and system for identifying the inner and outer boundary points of a point cloud contour, thereby solving the problem in existing technologies where point cloud boundary point extraction cannot distinguish between inner and outer boundary points.

[0005] This invention provides a method for identifying inscribed and circumscribed boundary points of a point cloud contour, comprising the following steps: S1. Perform planar segmentation on the original point cloud to obtain multiple independent planar point clouds; S2. For each planar point cloud, obtain its projection plane through planar fitting, extract the boundary points of the planar point cloud, and project the boundary points onto the corresponding projection plane; S3. Transform the projected 3D boundary point set to the xoy plane to obtain the 2D boundary point set; S4. Construct an envelope rectangle for the two-dimensional boundary point set using the minimum bounding rectangle algorithm; S5. Perform interval scanning on the two-dimensional boundary point set, calculate the distance from each two-dimensional boundary point within the scanning range to its corresponding rectangle edge, determine the two-dimensional boundary point with the smallest distance as the outer boundary point, and determine the remaining two-dimensional boundary points as the inner boundary points or redundant points.

[0006] Preferably, in S1, the original point cloud is segmented into planes using a region growing algorithm.

[0007] Preferably, in S2, the boundary points of the planar point cloud are extracted using a global constraint method based on the geometric feature distribution of neighboring points.

[0008] Preferably, in S2, the boundary points are projected onto the corresponding projection plane through spatial coordinate transformation: for each three-dimensional boundary point before projection, a spatial point projection rotation matrix is ​​first applied for rotation transformation, and then a spatial point projection translation matrix is ​​applied for translation transformation, thereby obtaining the coordinates of the three-dimensional boundary points after projection.

[0009] Preferably, in step S3, the projection plane normal vector and the unit vector in the z-axis direction are cross-multiplied to obtain a transformation direction feature vector, and the transformation direction feature vector is normalized to obtain a normalized transformation direction feature vector; the rotation angle is calculated based on the projection plane normal vector and the unit vector in the z-axis direction; a four-element matrix is ​​constructed based on the rotation angle and the normalized transformation direction feature vector; a three-dimensional coordinate system rotation matrix is ​​obtained based on the four-element matrix; and the coordinates of the projected three-dimensional boundary points are multiplied by the three-dimensional coordinate system rotation matrix to obtain the coordinates of the two-dimensional boundary points.

[0010] Preferably, in S5, the coordinates of a point on a certain side of the rectangle are used as the query point. Based on this query point and its corresponding rectangular edge vector The query point was obtained. Corresponding rectangle side line Determine the line connecting the sides of the rectangle on the xoy plane. Vertical scan lines .

[0011] Preferably, the remaining points after deleting the four inflection points of the rectangle are selected as candidate query points.

[0012] Preferably, the distance from each two-dimensional boundary point to the scan line is calculated. The distance is used to filter out the set of boundary points whose distance is less than the scanning distance threshold. ; Calculate the boundary point set The straight line from each 2D boundary point to the edge of the rectangle distance ; distance The smallest two-dimensional boundary point is determined to be a circumscribed boundary point.

[0013] Preferably, the scanning distance threshold is determined based on the original point cloud density.

[0014] On the other hand, the present invention provides a point cloud contour inscribed boundary point recognition system, comprising: Planar segmentation unit is used to perform planar segmentation on the original point cloud to obtain multiple independent planar point clouds; The boundary point extraction and projection unit is used to obtain the projection plane of each planar point cloud through planar fitting, extract the boundary points of the planar point cloud, and project the boundary points onto the corresponding projection plane. The coordinate transformation unit is used to transform the projected 3D boundary point set to the xoy plane to obtain a 2D boundary point set. Envelope rectangle building unit, used to construct an envelope rectangle from a set of two-dimensional boundary points using the minimum bounding rectangle algorithm; The inscribed and circumscribed boundary point discrimination unit is used to perform interval scanning of the two-dimensional boundary point set, calculate the distance from each boundary point within the scanning range to its corresponding rectangle edge, and determine the boundary point with the smallest distance as the circumscribed boundary point, and the remaining boundary points as inscribed boundary points or redundant points. The point cloud contour inscribed boundary point recognition system is used to perform the steps in the point cloud contour inscribed boundary point recognition method described above.

[0015] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention first performs planar segmentation on the original point cloud to obtain multiple independent planar point clouds. Then, for each planar point cloud, its projection plane is obtained through planar fitting. The boundary points of the planar point cloud are extracted and projected onto the corresponding projection plane. The projected 3D boundary point set is then transformed into the XY plane to obtain a 2D boundary point set. Next, the minimum bounding rectangle algorithm is used to construct an envelope rectangle for the 2D boundary point set. Finally, the 2D boundary point set is scanned at intervals, and the distance from each 2D boundary point within the scan range to its corresponding rectangle edge is calculated. The 2D boundary point with the smallest distance is identified as the circumscribed boundary point, and the remaining 2D boundary points are identified as inscribed boundary points or redundant points. Since this invention mainly utilizes the minimum bounding rectangle and scan distance constraints to identify the external boundary points (i.e., circumscribed boundary points) and internal boundary points (including inscribed boundary points or internal redundant points) of the point cloud contour, the problem of unidentified concave boundaries can be avoided. This invention, under the premise of the circumscribed rectangular envelope, makes full use of the distance constraint from the point to the corresponding rectangular edge in the scanning interval to achieve accurate identification of the inner and outer boundary points. It is also efficient and robust, and can meet the needs of rapid processing of large-scale point cloud data. Moreover, this invention does not involve too many parameters and redundant algorithms, which significantly improves the efficiency and refinement of the identification of the inner and outer boundary points of the point cloud contour. It is particularly suitable for the identification of the inner and outer boundary points of the surface contour of large scene point clouds. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for identifying inscribed and circumscribed boundary points of a point cloud contour, as provided in Embodiment 1 of the present invention.

[0017] Figure 2 This is an example diagram corresponding to the point cloud contour inscribed boundary point identification method provided in Embodiment 1 of the present invention; wherein, Figure 2 In the image, (a), (b), and (c) are the original point clouds corresponding to the three examples, respectively. Figure 2 In the diagram, (d), (e), and (f) represent the different planar point clouds segmented for the three examples. Figure 2 In the example, (g), (h), and (i) are the boundary points of different dividing planes corresponding to the three examples.

[0018] Figure 3 This is a schematic diagram illustrating a method for identifying the inner and outer boundary points of a point cloud contour according to Embodiment 1 of the present invention; wherein, Figure 3 (a) in the diagram is a schematic diagram of the simulated boundary points. Figure 3 (b) in the diagram is a schematic of the actual boundary points.

[0019] Figure 4 This is an image showing the recognition result of inscribed and circumscribed boundary points obtained using a point cloud contour inscribed and circumscribed boundary point recognition method provided in Embodiment 1 of the present invention; wherein, Figure 4Image (a) shows the identification results of the inner and outer boundary points for a portion of the region in Example 1. Figure 4 (b) in the figure shows the identification results of the inner and outer boundary points corresponding to a part of the region in Example 2. Detailed Implementation

[0020] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0021] Example 1: Example 1 provides a method for identifying the inscribed and circumscribed boundary points of a point cloud contour. See [link to example]. Figure 1 This includes the following steps: S1. Perform planar segmentation on the original point cloud to obtain multiple independent planar point clouds; S2. For each planar point cloud, obtain its projection plane through planar fitting, extract the boundary points of the planar point cloud, and project the boundary points onto the corresponding projection plane; S3. Transform the projected 3D boundary point set to the xoy plane to obtain the 2D boundary point set; S4. Construct an envelope rectangle for the two-dimensional boundary point set using the minimum bounding rectangle algorithm; S5. Perform interval scanning on the two-dimensional boundary point set, calculate the distance from each two-dimensional boundary point within the scanning range to its corresponding rectangle edge, determine the two-dimensional boundary point with the smallest distance as the outer boundary point, and determine the remaining two-dimensional boundary points as the inner boundary points or redundant points.

[0022] In essence, this invention first uses surface point clouds for planar fitting to obtain a projection plane. Based on a point-to-plane projection algorithm, the extracted boundary points are projected onto this plane, and the boundary points are then transformed onto the xoy plane. Next, the minimum bounding rectangle (MBR) method is used to obtain the bounding rectangle of the boundary points. Finally, the boundary points are scanned at intervals, and the distances from the boundary points within the scanning intervals to the edges of the bounding rectangles are used to determine the inner and outer boundary points. This invention is not only applicable to the identification of inner and outer boundaries of regular point clouds, but also demonstrates high accuracy and adaptability in the identification of complex irregular inner and outer boundaries. Furthermore, it can identify redundant points within the boundary clouds, providing an efficient, robust, flexible, and concise solution for the application requirements of fine-structure analysis of point cloud contours.

[0023] The steps of this invention will be described in detail below.

[0024] S1. Perform planar segmentation on the original point cloud to obtain multiple independent planar point clouds.

[0025] That is, the point cloud data is segmented into planes to obtain point cloud data in different planes.

[0026] Specifically, in S1, a region growing algorithm can be used to perform planar segmentation on the original point cloud. The effect of planar segmentation can be seen in [reference needed]. Figure 2 ,in, Figure 2 (a) in the image represents the original point cloud corresponding to the first example. Figure 2 In the example, (d) represents the different planar point clouds corresponding to the first example. Figure 2 (b) in the image represents the original point cloud corresponding to the second example. Figure 2 In the example, (e) represents the different planar point clouds corresponding to the second example. Figure 2 (c) in the image represents the original point cloud corresponding to the third example. Figure 2 In the example, (f) represents the different planar point clouds corresponding to the third example.

[0027] S2. For each planar point cloud, obtain its projection plane through planar fitting, extract the boundary points of the planar point cloud, and project the boundary points onto the corresponding projection plane.

[0028] That is, S2 includes two sub-steps: (1) Extract the boundary points of different surface point clouds obtained by S1.

[0029] (2) Perform plane fitting on the plane point cloud segmented by S1 to construct the projection plane.

[0030] In this invention, S2 can either perform plane fitting first and then extract boundary points, or extract boundary points first and then perform plane fitting.

[0031] Specifically, the boundary points of a planar point cloud can be extracted using a global constraint method based on the geometric feature distribution of neighboring points.

[0032] For the results of boundary point extraction, please refer to Figure 2 As shown in (g) to (i), where, Figure 2 In the example, (g) represents the boundary points of different dividing planes corresponding to the first example. Figure 2 In the example, (h) represents the boundary points of different dividing planes corresponding to the second example. Figure 2 In the example, (i) represents the boundary points of different dividing planes corresponding to the third example.

[0033] The boundary points are projected onto the corresponding projection plane by spatial coordinate transformation: For each 3D boundary point before projection, a spatial point projection rotation matrix is ​​first applied for rotation transformation, and then a spatial point projection translation matrix is ​​applied for translation transformation, so as to obtain the coordinates of the 3D boundary point after projection.

[0034] S3. Transform the projected 3D boundary point set to the xoy plane to obtain the 2D boundary point set.

[0035] Specifically, in S3, the cross product operation is performed between the projection plane normal vector and the unit vector in the z-axis direction to obtain the transformation direction feature vector, and the transformation direction feature vector is normalized to obtain the normalized transformation direction feature vector; the rotation angle is calculated based on the projection plane normal vector and the unit vector in the z-axis direction; a quartic matrix is ​​constructed based on the rotation angle and the normalized transformation direction feature vector; a three-dimensional coordinate system rotation matrix is ​​obtained based on the quartic matrix; and the coordinates of the projected three-dimensional boundary points are multiplied by the three-dimensional coordinate system rotation matrix to obtain the coordinates of the two-dimensional boundary points.

[0036] The following explanation uses the formulas to illustrate S2 and S3.

[0037] For the planar point cloud segmented by S1, a plane fitting equation is constructed as shown in equation (1): (1) In the formula, a, b, c, and d are plane fitting parameters.

[0038] The spatial point projection rotation matrix is ​​constructed according to equation (1) as follows: (2) In the formula, This is the rotation matrix for the projection of a point in space.

[0039] The spatial point projection translation matrix is: (3) In the formula, T is the spatial point projection translation matrix.

[0040] According to equations (2) and (3), the boundary points extracted in S2 can be projected onto the corresponding projection plane, as shown in equation (4): (4) In the formula, Let P be the three-dimensional boundary point after projection, and let P be the three-dimensional boundary point before projection.

[0041] Points on the projection plane are still three-dimensional points. To facilitate the identification of inner and outer boundary points, it is necessary to convert these points into two-dimensional points. This involves converting the projection plane normal vector... unit vector in the z-axis direction Perform the cross product, as shown in equation (5): (5) In the formula, The normal vector of the projection plane. Let u be the z-axis unit vector, and u be the transformation direction feature vector.

[0042] The normalization of equation (5) is shown in equation (6): (6) In the formula, The transformation direction feature vector is normalized.

[0043] The rotation angle is obtained using the four-element transformation method, as shown in equation (7): (7) In the formula, The angle is the rotation angle.

[0044] Based on equations (6) and (7), the four elements are obtained as follows: (8) In the formula, q represents four elements.

[0045] Based on these four elements, the rotation matrix is ​​obtained as follows: (9) In the formula, q(1), q(2), q(3), and q(4) are the first, second, third, and fourth elements of the four elements, respectively.

[0046] The coordinates of the spatial point in two dimensions are obtained according to equation (9), as shown in equation (10): (10) In the formula, These are the coordinates of the boundary point in two-dimensional space.

[0047] in, , .

[0048] because ,therefore, It can be in two-dimensional form Use it.

[0049] S4. Construct an envelope rectangle for the two-dimensional boundary point set using the minimum bounding rectangle algorithm.

[0050] This invention employs an existing minimum bounding rectangle algorithm to construct the envelope rectangle. The constructed minimum bounding rectangle can be found in [reference needed]. Figure 3 .

[0051] S5. Perform interval scanning on the two-dimensional boundary point set, calculate the distance from each two-dimensional boundary point within the scanning range to its corresponding rectangle edge, determine the two-dimensional boundary point with the smallest distance as the outer boundary point, and determine the remaining two-dimensional boundary points as the inner boundary points or redundant points.

[0052] For details, see Figure 3 ,in, Figure 3(a) in the diagram is a schematic diagram of the simulated boundary points. Figure 3 (b) in the diagram is a schematic diagram of the actual boundary points. In this invention, the coordinates of a point on a certain side of the rectangle are used as the query point. Based on this query point and its corresponding rectangular edge vector The query point was obtained. Corresponding rectangle side line Determine the line connecting the sides of the rectangle on the xoy plane. Vertical scan lines In this invention, the remaining points after deleting the four inflection points of the rectangle are preferably used as candidate query points to eliminate the influence of the four inflection points of the rectangle on the determination of the external boundary points, thereby obtaining a more accurate identification result.

[0053] Calculate the distance from each 2D boundary point to the scan line. The distance is used to filter out samples whose distance is less than the scanning distance threshold. Boundary point set ; Calculate the boundary point set The straight line from each 2D boundary point to the edge of the rectangle distance ; distance The smallest two-dimensional boundary point is determined as the circumscribed boundary point (also denoted as the surface boundary point), while other boundary points are determined as inscribed boundary points or redundant points. The scanning distance threshold is mentioned below. It can be determined based on the original point cloud density.

[0054] The following explanation uses the formula as an example.

[0055] This invention can denote the four rectangular side vectors of the circumscribed rectangle as: (j=1,2,3,4), mark the coordinates of any point on the corresponding edge as follows: (j=1,2,3,4). Taking a certain rectangle edge as an example, based on each query point on the circumscribed rectangle edge... and the corresponding rectangle edge vector This yields the straight line of the rectangle corresponding to the query point. Rectangle side straight line It is expressed as follows: (11) in, , .

[0056] Determine the straight line between the rectangle and the xoy plane. Perpendicular vectors Based on the query point and vector , obtain the scan line , scan line It is expressed as follows: (12) in, .

[0057] Then calculate all boundary points to the sweep line. The distance (from the i-th boundary point to) The distance is denoted as ), and obtain the result by satisfying the conditions The set of boundary points As shown in equation (13).

[0058] (13) In the formula, Boundary point To scan line Shortest distance, For the boundary point, Represents the real number field.

[0059] Then calculate the set of boundary points within this interval. Each boundary point in the middle (the set of boundary points) The i-th boundary point in the equation is denoted as . ( ) to the straight line of the rectangle side distance ( arrive The distance is denoted as ),like Figure 3 As shown in (a) above. Based on distance The minimum value is used to determine whether each boundary point belongs to an external boundary point, an internal boundary point, or a redundant point, as shown in equation (14).

[0060] (14) Finally, equation (14) is used to identify the inscribed and circumscribed boundary points of the point cloud contour. See the identification results below. Figure 4 Since this invention does not specifically distinguish between inscribed boundary points and redundant points, for the sake of simplicity, in Figure 4 In this context, all inscribed boundary points or redundant points are collectively marked as inscribed boundary points. Figure 4 Image (a) shows the identification results of the inner and outer boundary points for a portion of the region in Example 1. Figure 4 (b) in the figure shows the identification results of the inner and outer boundary points corresponding to a part of the region in Example 2.

[0061] In summary, this invention addresses the specific problem of identifying circumscribed and inscribed boundary points in point cloud contour design. It utilizes the minimum bounding rectangle and scan distance constraints to identify circumscribed and inscribed boundary points or internal redundant points of point cloud contours, avoiding the problem of unidentified concave boundaries. Under the premise of the circumscribed rectangle envelope, this invention fully leverages the distance constraints from points in the scan interval to the corresponding rectangle edges to achieve accurate identification of circumscribed and inscribed boundary points. It also possesses high efficiency and robustness, meeting the rapid processing needs of large-scale point cloud data. Furthermore, this invention involves minimal parameters and redundant algorithms, significantly improving the efficiency and precision of point cloud contour circumscribed and inscribed boundary point identification, making it particularly suitable for identifying circumscribed and inscribed boundary points of large-scale point cloud surface contours.

[0062] Furthermore, because this invention can quickly identify circumscribed and concentric contour points without relying on large amounts of high-quality labeled data and parameters, and the identified circumscribed and concentric contour points are more detailed, the precision of point cloud contour point identification using this invention is superior to α-shape, convex hull, or deep learning methods. This invention's method is also more robust than the latter methods and less susceptible to point cloud noise.

[0063] Example 2: Example 2 provides a point cloud contour inscribed boundary point recognition system, including: Planar segmentation unit is used to perform planar segmentation on the original point cloud to obtain multiple independent planar point clouds; The boundary point extraction and projection unit is used to obtain the projection plane of each planar point cloud through planar fitting, extract the boundary points of the planar point cloud, and project the boundary points onto the corresponding projection plane. The coordinate transformation unit is used to transform the projected 3D boundary point set to the xoy plane to obtain a 2D boundary point set. Envelope rectangle building unit, used to construct an envelope rectangle from a set of two-dimensional boundary points using the minimum bounding rectangle algorithm; The inscribed / extra-boundary point discrimination unit is used to perform interval scanning of the two-dimensional boundary point set, calculate the distance from each boundary point within the scanning range to its corresponding rectangular edge, and determine the boundary point with the smallest distance as the inscribed boundary point, and the remaining boundary points as inscribed boundary points or redundant points.

[0064] The point cloud contour inscribed boundary point recognition system provided in Example 2 is used to perform the steps in the point cloud contour inscribed boundary point recognition method as described in Example 1.

[0065] Since the functions of each unit in the point cloud contour inscribed boundary point recognition system provided in Embodiment 2 correspond to the steps in the point cloud contour inscribed boundary point recognition method provided in Embodiment 1, Embodiment 2 can be understood by referring to the description of Embodiment 1, and will not be repeated here.

[0066] Finally, it should be noted that the above specific embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying inscribed and circumscribed boundary points of a point cloud contour, characterized in that, Includes the following steps: S1. Perform planar segmentation on the original point cloud to obtain multiple independent planar point clouds; S2. For each planar point cloud, obtain its projection plane through planar fitting, extract the boundary points of the planar point cloud, and project the boundary points onto the corresponding projection plane; S3. Transform the projected 3D boundary point set to the xoy plane to obtain the 2D boundary point set; S4. Construct an envelope rectangle for the two-dimensional boundary point set using the minimum bounding rectangle algorithm; S5. Perform interval scanning on the two-dimensional boundary point set, calculate the distance from each two-dimensional boundary point within the scanning range to its corresponding rectangle edge, determine the two-dimensional boundary point with the smallest distance as the outer boundary point, and determine the remaining two-dimensional boundary points as the inner boundary points or redundant points. In S5, the coordinates of a point on one side of a rectangle are used as the query point. Based on this query point and its corresponding rectangular edge vector The query point was obtained. Corresponding rectangle side line Determine the line connecting the rectangle's sides on the xoy plane. Vertical scan lines ; The remaining points after deleting the four inflection points of the rectangle will be used as candidate query points; Calculate the distance from each 2D boundary point to the scan line. The distance is used to filter out the set of boundary points whose distance is less than the scanning distance threshold. ; Calculate the boundary point set The straight line from each 2D boundary point to the edge of the rectangle distance ; distance The smallest two-dimensional boundary point is determined to be a circumscribed boundary point.

2. The point cloud contour inscribed boundary point identification method according to claim 1, characterized in that, In S1, the original point cloud is segmented into planes using a region growing algorithm.

3. The point cloud contour inscribed boundary point identification method according to claim 1, characterized in that, In S2, the boundary points of the planar point cloud are extracted using a global constraint method based on the geometric feature distribution of neighboring points.

4. The point cloud contour inscribed boundary point identification method according to claim 1, characterized in that, In S2, the boundary points are projected onto the corresponding projection plane through spatial coordinate transformation: for each three-dimensional boundary point before projection, a spatial point projection rotation matrix is ​​first applied for rotation transformation, and then a spatial point projection translation matrix is ​​applied for translation transformation, thereby obtaining the coordinates of the three-dimensional boundary points after projection.

5. The point cloud contour inscribed boundary point identification method according to claim 1, characterized in that, In step S3, the cross product operation is performed between the projection plane normal vector and the unit vector in the z-axis direction to obtain the transformation direction feature vector. The transformation direction feature vector is then normalized to obtain a normalized transformation direction feature vector. The rotation angle is calculated based on the projection plane normal vector and the unit vector in the z-axis direction. A quaternion is constructed based on the rotation angle and the normalized transformation direction feature vector. A three-dimensional coordinate system rotation matrix is ​​obtained based on the quaternion. The coordinates of the projected three-dimensional boundary points are multiplied by the three-dimensional coordinate system rotation matrix to obtain the coordinates of the two-dimensional boundary points.

6. The method for identifying inscribed and circumscribed boundary points of a point cloud contour according to claim 1, characterized in that, The scanning distance threshold is determined based on the original point cloud density.

7. A point cloud contour inscribed boundary point recognition system, characterized in that, include: Planar segmentation unit is used to perform planar segmentation on the original point cloud to obtain multiple independent planar point clouds; The boundary point extraction and projection unit is used to obtain the projection plane of each planar point cloud through planar fitting, extract the boundary points of the planar point cloud, and project the boundary points onto the corresponding projection plane. The coordinate transformation unit is used to transform the projected 3D boundary point set to the xoy plane to obtain a 2D boundary point set. Envelope rectangle building unit, used to construct an envelope rectangle from a set of two-dimensional boundary points using the minimum bounding rectangle algorithm; The inscribed and circumscribed boundary point discrimination unit is used to perform interval scanning of the two-dimensional boundary point set, calculate the distance from each boundary point within the scanning range to its corresponding rectangle edge, and determine the boundary point with the smallest distance as the circumscribed boundary point, and the remaining boundary points as inscribed boundary points or redundant points. The point cloud contour inscribed boundary point recognition system is used to perform the steps in the point cloud contour inscribed boundary point recognition method as described in any one of claims 1 to 6.

Images