Intelligent positioning method and system for hyperbolic plate processing

Abstract

The invention relates to the field of image processing, in particular to an intelligent positioning method and system for hyperbolic plate processing. Comprising the following steps: acquiring a three-dimensional scanning model and a theoretical three-dimensional model of the hyperbolic plate, and acquiring a scanning topological graph and a theoretical topological graph; obtaining a scanning priority order and a theoretical priority order; obtaining an optimal matching point of each vertex in the scanning topological graph in the theoretical topological graph, and determining a theoretical point corresponding to each optimal matching point; respectively determining an original point corresponding to each vertex in the scanning topological graph; forming a three-dimensional point corresponding sequence by the original points and the theoretical points corresponding to the vertexes in each scanning topological graph, and obtaining a mapping relation corresponding to the three-dimensional point corresponding sequence through an SIFT method; obtaining a precise registration transformation matrix; and decomposing the fine registration transformation matrix to obtain a variable quantity, converting the variable quantity into a control instruction, and carrying out positioning adjustment on a defect link in the hyperbolic plate manufacturing process. The production precision of the hyperbolic plate can be improved.

Description

Technical Field

[0001] This invention relates to the field of image processing, and more specifically to an intelligent positioning method and system for hyperboloid plate processing. Background Technology

[0002] When performing high-precision intelligent positioning for hyperbolic plate processing, it is first necessary to obtain its actual three-dimensional point cloud data through laser scanning, and then use the ICP (Iterative Closest Point) algorithm to accurately match and compare the scanned data with the theoretical CAD design model, thereby calculating the accurate pose deviation for subsequent processing correction.

[0003] However, the unique smooth, continuous, sparse, and gently curvature-changing geometry of the hyperbolic plate causes the core "nearest point search" step of the algorithm to lack a clear corresponding target, resulting in a large number of fuzzy and erroneous matching point pairs on the surface. At the same time, the similar local geometric features between different regions make the iterative process highly dependent on the initial pose, which easily gets trapped in local optima and fails to converge to the globally correct registration position, severely restricting the final matching accuracy and reliability. Summary of the Invention

[0004] This invention provides an intelligent positioning method and system for hyperboloid plate processing to solve existing problems.

[0005] The present invention provides an intelligent positioning method for hyperboloid plate processing, which adopts the following technical solution: One embodiment of the present invention provides an intelligent positioning method for processing hyperboloid plates, the method comprising the following steps: Obtain the 3D scanning model and theoretical 3D model of the hyperbolic plate, and obtain the scanning topology diagram and theoretical topology diagram based on the 3D scanning model and theoretical 3D model; Obtain the topological relationship strength of each vertex in the scanned topological graph and the topological relationship strength of each vertex in the theoretical topological graph, and sort them respectively to obtain the scan priority sort and the theoretical priority sort. Based on the scanning priority sorting and theoretical priority sorting, obtain the optimal matching point of each vertex in the scanning topology graph in the theoretical topology graph, and determine the theoretical point corresponding to each optimal matching point; Determine the original point corresponding to each vertex in the scanned topology graph; The original points and theoretical points corresponding to each vertex in the scanned topology graph are combined into a 3D point correspondence sequence, and the mapping relationship corresponding to the 3D point correspondence sequence is obtained by using the SIFT method. Based on the mapping relationship, the 3D scanning model, and the theoretical 3D model, obtain the fine registration transformation matrix; The fine registration transformation matrix is ​​decomposed to obtain the change quantity, which is then converted into control commands to locate and adjust the defective links in the hyperbolic plate manufacturing process.

[0006] Optionally, obtaining the scan topology map and theoretical topology map based on the 3D scan model and theoretical 3D model specifically includes: Viewpoint and depth maps are obtained from the 3D scanning model and the theoretical 3D model, respectively, with the same parameters for obtaining the viewpoint and depth maps. The SIFT algorithm is used to obtain the scan key points and theoretical key points corresponding to the view map and depth map, respectively. Using the Delaunay triangulation method, triangulation networks are constructed with scanned key points and theoretical key points as vertices, respectively, to obtain scanned topology and theoretical topology.

[0007] Optionally, obtaining the topological relation strength of each vertex in the scanned topological graph and the topological relation strength of each vertex in the theoretical topological graph specifically includes: Obtain the degree of each vertex in the topological graph, and determine the relative degree centrality of the Nth vertex by the ratio of the degree of the Nth vertex to the maximum degree of the vertex. The topological graph is either a scanned topological graph or a theoretical topological graph. The distances between the Nth vertex and other vertices in the topological graph are obtained by using Dijkstra's shortest path algorithm and summed to obtain the distance sum of the Nth vertex. The negative number of the distance sum of the Nth vertex is then used as the exponent of the natural base to obtain the centrality of the Nth vertex. The product of the relative degree centrality of the Nth vertex and the centrality of the Nth vertex is determined as the topological relation strength of the Nth vertex; Determine the topological relation strength of each vertex in the scanned topological graph and the topological relation strength of each vertex in the theoretical topological graph, respectively.

[0008] Optionally, obtaining the optimal matching point of each vertex in the scanned topology graph in the theoretical topology graph based on scan priority sorting and theoretical priority sorting specifically includes: The vertex corresponding to the first element in the scan priority sorting in the scan topology graph is determined as the first matching point; Based on the first matching point, candidate matching points corresponding to the first matching point are determined from the theoretical priority ranking, wherein the absolute value of the difference between the degree of the first matching point and the degree of the candidate matching point is less than a preset degree threshold. Calculate the similarity between the descriptor of each candidate matching point and the descriptor of the first matching point, and determine the candidate matching points with similarity greater than a preset similarity threshold as retained candidate matching points; Determine the priority sequence of neighboring points of the first matching point based on the scanned topology map; Determine the neighborhood priority sequence of candidate matching points to be retained based on the theoretical topology graph; Obtain the cosine similarity between the scanned topology graph and the theoretical topology graph; The matching degree between the first matching point and each retained candidate matching point is calculated based on the cosine similarity. The retained candidate matching point with the highest matching degree is determined as the optimal matching point of the first matching point. The vertex corresponding to the Ath element in the scan priority sorting in the scan topology graph is determined as the Ath matching point; Obtain the optimal matching point of the vertex corresponding to each element in the scan topology graph in the scan priority sorting.

[0009] Optionally, determining the priority sequence of neighboring points of the first matching point based on the scanned topology map specifically includes: The neighborhood points of the first matching point are determined from the scanned topology graph, wherein the neighborhood points of the first matching point and the first matching point are connected in the scanned topology graph through the edges of the scanned topology graph; Calculate the similarity between the descriptors of the neighboring points of each first matching point and the descriptor of the first matching point, and sort them in ascending order to obtain the sequence of neighboring points of the first matching point; The index value of each element in the scan priority sort is determined as the priority sort value of the corresponding vertex in the scan topology graph, thus obtaining the priority sort value of each vertex in the scan topology graph; The value of each element in the neighborhood point sequence of the first matching point is replaced with the priority sorting value of the corresponding vertex in the scanned topology graph, to obtain the neighborhood point priority sequence.

[0010] Optionally, determining the neighborhood priority sequence of candidate matching points based on the theoretical topology graph specifically includes: The neighborhood points of the candidate matching points are determined from the theoretical topology graph, wherein the neighborhood points of the candidate matching points and the candidate matching points are connected by edges in the theoretical topology graph. Calculate the similarity between the descriptors of the neighboring points of each retained candidate matching point and the descriptors of the retained candidate matching points, and sort them in ascending order to obtain the sequence of neighboring points of the retained candidate matching points; The index value of each element in the theoretical priority sort is determined as the priority sort value of the corresponding vertex in the theoretical topology graph, thus obtaining the priority sort value of each vertex in the theoretical topology graph. The value of each element in the neighborhood point sequence of the retained candidate matching point is replaced with the priority ranking value of the corresponding vertex in the theoretical topology graph, thus obtaining the neighborhood priority sequence.

[0011] Optionally, obtaining the cosine similarity between the scanned topology graph and the theoretical topology graph specifically includes: Calculate the eigenvalues ​​of the normalized Laplacian matrix of the scanned topology graph, and sort the eigenvalues ​​in descending order to obtain the first eigenvalue sequence; Calculate the eigenvalues ​​of the normalized Laplacian matrix of the theoretical topology graph, and arrange the eigenvalues ​​in descending order to obtain the second eigenvalue sequence; Obtain the elements in the first feature value sequence that are greater than the preset first feature value to obtain the first vector; Obtain the elements in the second feature value sequence that are greater than the preset second feature value to obtain the second vector; Calculate the cosine similarity between the first vector and the second vector, and determine it as the cosine similarity between the scanned topological graph and the theoretical topological graph.

[0012] Optionally, the step of calculating the matching degree between the first matching point and each retained candidate matching point based on cosine similarity specifically includes: The similarity between the descriptor of the Mth retained candidate matching point and the descriptor of the first matching point is determined as the similarity of the Mth descriptor; The cosine similarity is negative, then 1 is added, and multiplied with the similarity of the Mth descriptor to obtain the first matching degree of the Mth retained candidate matching point; Calculate the similarity between the priority sequence of neighboring points and the priority sequence of the neighboring points of the Mth retained candidate matching point to obtain the similarity of the Mth neighborhood. The product of the similarity of the Mth neighborhood and the cosine similarity is determined as the second matching degree of the Mth retained candidate matching point; The product of the first matching degree and the second matching degree of the Mth retained candidate matching point is determined as the matching degree between the first matching point and the Mth retained candidate matching point. Obtain the matching degree between the first matching point and each retained candidate matching point.

[0013] Optionally, obtaining the fine registration transformation matrix based on the mapping relationship, the 3D scanning model, and the theoretical 3D model specifically includes: The mapping relationship, the preset minimum point set size, the interior point distance threshold, and the maximum number of iterations are used as inputs to the RANSAC robust estimation algorithm, which outputs the rigid body transformation matrix. The rigid body transformation matrix, the 3D scanning model, and the theoretical 3D model are input into the ICP algorithm, and the ICP algorithm outputs the fine registration transformation matrix.

[0014] This invention proposes an intelligent positioning system for hyperboloid plate machining, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the intelligent positioning method for hyperboloid plate machining as described above.

[0015] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, a depth map-based SIFT matching method is used to first generate multi-view depth maps from the hyperbolic slab scan point cloud and CAD model, and extract robust SIFT features that are invariant to rotation and scaling. Then, through high-dimensional descriptor matching and robust RANSAC estimation, a high-precision initial rigid body transformation matrix is ​​calculated, thereby achieving global coarse registration between the scan data and the design model. This effectively overcomes the matching ambiguity problem caused by the smooth surface and sparse features of the hyperbolic slab, providing a near-ideal iterative starting point for subsequent ICP matching. This allows it to focus on optimizing small local deformations without consuming iterative resources in incorrect directions, thus significantly improving the accuracy, convergence speed, and overall robustness of the final ICP matching. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart of an intelligent positioning method for hyperboloid plate processing provided in one embodiment of the present invention; Figure 2 A scatter plot of key points; Figure 3 This is a triangulation diagram obtained with key points as vertices; Figure 4 This is a structural diagram of an intelligent positioning system for hyperboloid plate processing provided in one embodiment of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent positioning method for hyperboloid plate processing proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details a specific scheme for an intelligent positioning method for hyperboloid plate processing provided by the present invention.

[0021] This invention provides an intelligent positioning method and system for hyperboloid plate processing. Please refer to [link / reference]. Figure 1 The diagram illustrates a flowchart of an intelligent positioning method for hyperboloid plate processing according to an embodiment of the present invention, the method comprising the following steps: S101. Obtain the 3D scanning model and theoretical 3D model of the hyperbolic plate, and obtain the scanning topology diagram and theoretical topology diagram based on the 3D scanning model and theoretical 3D model.

[0022] In this embodiment, obtaining the scan topology map and the theoretical topology map based on the 3D scan model and the theoretical 3D model specifically includes: Viewpoint and depth maps are obtained from the 3D scanning model and the theoretical 3D model, respectively, with the same parameters for obtaining the viewpoint and depth maps. The SIFT algorithm is used to obtain the scan key points and theoretical key points corresponding to the view map and depth map, respectively. Using the Delaunay triangulation method, triangulation networks are constructed with scanned key points and theoretical key points as vertices, respectively, to obtain scanned topology and theoretical topology.

[0023] For example, this process aims to comprehensively acquire the three-dimensional geometric information of the actual surface of the hyperboloid plate using precision measuring equipment and convert it into a digital model.

[0024] Data Acquisition: A high-precision laser scanner is used to perform a full-coverage scan of the hyperboloid workpiece surface. By planning the scanning path and leveraging the device's high-density sampling capabilities, a massive amount of 3D point coordinate data of the entire outer surface of the workpiece is acquired. During scanning, special reference markers are typically placed around the workpiece to provide a unified coordinate reference frame for the entire scanning process, ensuring global data consistency.

[0025] Data Fusion and Generation: The raw point cloud data acquired by the scanner is fused, denoised, and encapsulated using specialized processing software to ultimately generate a 3D scanned model that accurately reflects the actual manufacturing morphology of the hyperbolic plate (usually stored as triangular mesh patches, such as .stl or .obj files). This model fully records the actual dimensions and shape of the workpiece.

[0026] The process of obtaining the theoretical 3D model aims to obtain an accurate theoretical design model of the hyperboloid under ideal conditions: Data Source: This model originates directly from the initial design phase of the product. Designers used computer-aided design (CAD) software (such as CATIA, Siemens NX, etc.) to directly create a theoretical 3D model of the hyperbolic plate based on the product's aerodynamic, structural, and other performance requirements.

[0027] Data preparation: Extract the final CAD design file of the part from the Product Data Management (PDM) system. Depending on the needs of subsequent matching processes, the file may be converted to a different format (e.g., exported as .stp or .iges), but its geometric definitions and dimensional accuracy must be strictly guaranteed to be completely consistent with the original design.

[0028] For a 3D scanned model obtained through laser scanning, it can be projected onto a virtual image plane from six main view directions of the workpiece (up, down, left, right, front, and back; the specific angles of the main views can be adaptively added and adjusted by the user according to the actual shape of the hyperbolic plate), thus obtaining six perspective maps. The grayscale value of each pixel in each perspective map represents the distance from the corresponding point to the viewpoint.

[0029] For the theoretical 3D model, the CAD model is virtually rendered from the same perspective and parameters (focal length, image size, etc.) as the 3D scanning model, generating a depth map for each perspective (each of the 6 perspectives).

[0030] Thus, the 3D matching problem is transformed into a 2D image matching problem between multiple depth maps.

[0031] For the view map and depth map under the same viewpoint, keypoints in the image are obtained using the SIFT algorithm (keypoints in the view map are view map keypoints, and keypoints in the depth map are theoretical keypoints). Each keypoint corresponds to a 128-dimensional vector used to describe the position, scale, and orientation information of the keypoint.

[0032] Traditional methods directly select the keypoint with the highest vector similarity as the keypoint pair. However, due to the unique smooth and continuous, sparse feature and gently changing curvature geometry of the hyperbolic plate, multiple matching keypoints with high similarity and little difference can be obtained for each keypoint. If the matching keypoint with the highest similarity is directly selected as the matching point pair for that keypoint, a large number of fuzzy and erroneous matching point pairs are likely to be generated.

[0033] Based on this, this embodiment selects matching key points by combining more global information, considering which key point in which image should be prioritized for matching point calculation, and then obtaining the optimal matching point from among many matching points by combining the global information of the priority key point, rather than directly using the matching point with the highest similarity as the matching point of the priority key point in another image.

[0034] For each graph, a triangulation network is constructed using these key points as vertices through Delaunay triangulation to obtain a topological graph (which includes a scanned topological graph and a theoretical topological graph).

[0035] like Figure 2 As shown, Figure 2 A scatter plot of key points. Figure 2 The scattered points can be keypoints obtained using the SIFT algorithm from any viewpoint map or any depth map. For example... Figure 3 As shown, Figure 3 This is a triangulation diagram obtained with key points as vertices.

[0036] In this embodiment, points with stronger topological relationships on the topology graph are given priority for matching. These points are closely connected with other key points, like core hubs in the network, which can provide stronger constraints and higher reliability for the subsequent matching process, thereby effectively reducing the propagation of incorrect matches and improving the overall registration accuracy and robustness.

[0037] S102. Obtain the topological relationship strength of each vertex in the scanned topological graph and the topological relationship strength of each vertex in the theoretical topological graph, and sort them respectively to obtain the scan priority sort and the theoretical priority sort.

[0038] In this embodiment, obtaining the topological relationship strength of each vertex in the scanned topological graph and the topological relationship strength of each vertex in the theoretical topological graph specifically includes: Obtain the degree of each vertex in the topological graph, and determine the relative degree centrality of the Nth vertex by the ratio of the degree of the Nth vertex to the maximum degree of the vertex. The topological graph is either a scanned topological graph or a theoretical topological graph. The distances between the Nth vertex and other vertices in the topological graph are obtained by using Dijkstra's shortest path algorithm and summed to obtain the distance sum of the Nth vertex. The negative number of the distance sum of the Nth vertex is then used as the exponent of the natural base to obtain the centrality of the Nth vertex. The product of the relative degree centrality of the Nth vertex and the centrality of the Nth vertex is determined as the topological relation strength of the Nth vertex; Determine the topological relation strength of each vertex in the scanned topological graph and the topological relation strength of each vertex in the theoretical topological graph, respectively.

[0039] For example, taking the topology graph corresponding to any view or depth map as an example, for each vertex in the topology graph, the more edges directly connected to that vertex, the more "direct neighbors" it has. It's like a busy transportation hub, directly connecting many places, at least the center of a local area; the smaller its "total distance" to all other points in the graph, the less it needs to go through intermediate nodes to reach any place. Such vertices are closer to the true geometric center or information propagation center. Therefore, this embodiment quantifies the strength of the topological relationship of each vertex by using its degree and the reciprocal of the average shortest path length from each vertex to all other vertices in the graph.

[0040] For each vertex in each topological graph, the ratio of that vertex's degree to the maximum degree among all vertices is denoted as n1 (i.e., relative degree centrality); the distance from that vertex to every other vertex is obtained using Dijkstra's shortest path algorithm, yielding the distance and x; x is then... z is the centrality of this vertex.

[0041] For each vertex in each topological graph, the product of n1 and z of that vertex is taken as the topological relation strength of that vertex. The stronger the topological relation, the closer the connection with other key points, and the stronger the constraints and higher reliability it can provide for the subsequent matching process, thereby effectively reducing the propagation of erroneous matches.

[0042] For the view map and depth map under the same viewpoint, obtain the topological relationship strength of each key point, and then obtain the priority ranking of key points according to the order of the topological relationship strength of the key points from large to small (sorting the topological relationship strength of each vertex in the scan topological map to obtain the scan priority ranking; sorting the topological relationship strength of each vertex in the theoretical topological map to obtain the theoretical priority ranking), the earlier the key point in the sequence, the greater the matching priority.

[0043] In this embodiment, the sorting method can be either sequential or reverse order. Keypoints are sorted from strongest to weakest in the topological relationship strength, with higher matching priority for earlier keypoints in the sequence. Conversely, keypoints are sorted from weakest to strongest in the topological relationship strength, with higher matching priority for later keypoints in the sequence.

[0044] S103. Based on the scanning priority sorting and theoretical priority sorting, obtain the optimal matching point of each vertex in the scanning topology graph in the theoretical topology graph, and determine the theoretical point corresponding to each optimal matching point.

[0045] In this embodiment, the optimal matching point of each vertex in the scanned topology graph in the theoretical topology graph is obtained according to the scan priority sorting and the theoretical priority sorting, specifically including: The vertex corresponding to the first element in the scan priority sorting in the scan topology graph is determined as the first matching point; Based on the first matching point, candidate matching points corresponding to the first matching point are determined from the theoretical priority ranking, wherein the absolute value of the difference between the degree of the first matching point and the degree of the candidate matching point is less than a preset degree threshold. Calculate the similarity between the descriptor of each candidate matching point and the descriptor of the first matching point, and determine the candidate matching points with similarity greater than a preset similarity threshold as retained candidate matching points; Determine the priority sequence of neighboring points of the first matching point based on the scanned topology map; Determine the neighborhood priority sequence of candidate matching points to be retained based on the theoretical topology graph; Obtain the cosine similarity between the scanned topology graph and the theoretical topology graph; The matching degree between the first matching point and each retained candidate matching point is calculated based on the cosine similarity. The retained candidate matching point with the highest matching degree is determined as the optimal matching point of the first matching point. The vertex corresponding to the Ath element in the scan priority sorting in the scan topology graph is determined as the Ath matching point; Obtain the optimal matching point of the vertex corresponding to each element in the scan topology graph in the scan priority sorting.

[0046] The priority sequence of neighboring points of the first matching point is determined based on the scanned topology map, specifically including: The neighborhood points of the first matching point are determined from the scanned topology graph, wherein the neighborhood points of the first matching point and the first matching point are connected in the scanned topology graph through the edges of the scanned topology graph; Calculate the similarity between the descriptors of the neighboring points of each first matching point and the descriptor of the first matching point, and sort them in ascending order to obtain the sequence of neighboring points of the first matching point; The index value of each element in the scan priority sort is determined as the priority sort value of the corresponding vertex in the scan topology graph, thus obtaining the priority sort value of each vertex in the scan topology graph; The value of each element in the neighborhood point sequence of the first matching point is replaced with the priority sorting value of the corresponding vertex in the scanned topology graph, to obtain the neighborhood point priority sequence.

[0047] The neighborhood priority sequence for retaining candidate matching points is determined based on the theoretical topology graph, specifically including: The neighborhood points of the candidate matching points are determined from the theoretical topology graph, wherein the neighborhood points of the candidate matching points and the candidate matching points are connected by edges in the theoretical topology graph. Calculate the similarity between the descriptors of the neighboring points of each retained candidate matching point and the descriptors of the retained candidate matching points, and sort them in ascending order to obtain the sequence of neighboring points of the retained candidate matching points; The index value of each element in the theoretical priority sort is determined as the priority sort value of the corresponding vertex in the theoretical topology graph, thus obtaining the priority sort value of each vertex in the theoretical topology graph. The value of each element in the neighborhood point sequence of the retained candidate matching point is replaced with the priority ranking value of the corresponding vertex in the theoretical topology graph, thus obtaining the neighborhood priority sequence.

[0048] Obtain the cosine similarity between the scanned topology graph and the theoretical topology graph, specifically including: Calculate the eigenvalues ​​of the normalized Laplacian matrix of the scanned topology graph, and sort the eigenvalues ​​in descending order to obtain the first eigenvalue sequence; Calculate the eigenvalues ​​of the normalized Laplacian matrix of the theoretical topology graph, and arrange the eigenvalues ​​in descending order to obtain the second eigenvalue sequence; Obtain the elements in the first feature value sequence that are greater than the preset first feature value to obtain the first vector; Obtain the elements in the second feature value sequence that are greater than the preset second feature value to obtain the second vector; Calculate the cosine similarity between the first vector and the second vector, and determine it as the cosine similarity between the scanned topological graph and the theoretical topological graph.

[0049] The matching degree between the first matching point and each retained candidate matching point is calculated based on cosine similarity, specifically including: The similarity between the descriptor of the Mth retained candidate matching point and the descriptor of the first matching point is determined as the similarity of the Mth descriptor; The cosine similarity is negative, then 1 is added, and multiplied with the similarity of the Mth descriptor to obtain the first matching degree of the Mth retained candidate matching point; Calculate the similarity between the priority sequence of neighboring points and the priority sequence of the neighboring points of the Mth retained candidate matching point to obtain the similarity of the Mth neighborhood. The product of the similarity of the Mth neighborhood and the cosine similarity is determined as the second matching degree of the Mth retained candidate matching point; The product of the first matching degree and the second matching degree of the Mth retained candidate matching point is determined as the matching degree between the first matching point and the Mth retained candidate matching point. Obtain the matching degree between the first matching point and each retained candidate matching point.

[0050] For example, taking both scan priority sorting and scan priority ranking as sorting from largest to smallest as an example. Obtain the first keypoint in the scan priority ranking, i.e., the first matching point, and denote this matching point as the first matching point. Represent the degree of the first matching point as i. Obtain the degree of each keypoint in the theoretical priority ranking, and select keypoints whose absolute difference from i is less than 3 as candidate matching points. For each candidate matching point, first calculate the similarity between each candidate matching point and the descriptor of the first matching point. Retain candidate matching points with a similarity greater than 0.7, and denote them as retained candidate matching points.

[0051] Optionally, the preset degree threshold and preset similarity threshold are 3 and 0.7 respectively. This is just an example, and users can adjust them according to actual production needs. No specific numerical restrictions are imposed here.

[0052] In this embodiment, the coherence of features between the matching point and its neighboring points is used as an important indicator to measure its similarity with the candidate point. That is, if a candidate point has a high feature similarity with the point to be matched, and the neighborhood relationship and feature distribution reflected by its surrounding topology are also highly consistent with the relationship between the point to be matched and its neighborhood, then the matching priority of the candidate point is higher.

[0053] For the first matching point, the key points directly connected to the point on the scan topology map are recorded as the neighboring points of the point. The similarity between each neighboring point and the descriptor of the point is calculated, and the similarity is sorted in ascending order to obtain the corresponding neighboring point sequence.

[0054] Obtain the topological relationship strength of each vertex on the scanned topological graph where the first matching point is located. According to the order of strength from largest to smallest, obtain the priority ranking value of each vertex. For example, the priority ranking value of the vertex with the largest topological relationship strength is 1, the priority ranking value of the vertex with the second largest topological relationship strength is 2, and so on.

[0055] For the first matching point, the label of each neighboring point in the neighboring point sequence is represented by the priority sorting value of the corresponding point to obtain the neighboring point priority sequence.

[0056] The construction of the neighborhood point priority sequence involves two levels of information representation: First, the sequence is ranked based on the descriptor similarity between neighborhood points and the first matching point, reflecting the feature similarity relationship between each neighborhood point and the first matching point; second, by replacing the similarity values ​​with the corresponding topological priority ranking values, the sequence further expresses the relative importance distribution among neighborhood points based on the strength of topological connections. Therefore, the neighborhood point priority sequence not only characterizes the association properties between the first matching point and its neighborhood points, but also reveals the structural salience of nodes at different positions in the local neighborhood in a quantitative manner.

[0057] For each retained candidate matching point of the first matching point, the neighborhood point sequence of each retained candidate matching point is obtained by the same method. The topological relationship strength of each key point on the theoretical topology graph where the retained candidate matching point is located is obtained, and then the neighborhood priority sequence is obtained.

[0058] For the scanned topology of the image where the first matching point is located, the eigenvalues ​​of the normalized Laplacian matrix of the scanned topology are calculated, and the first eigenvalue sequence is obtained in descending order. For each retained candidate matching point, the second eigenvalue sequence of the corresponding theoretical topology matrix is ​​also calculated using the same method. The first 7 largest eigenvalues ​​of each eigenvalue sequence are extracted (the specific number of extractions can be set by the user according to their needs) to form two 7-dimensional vectors. The cosine similarity s of the two 7-dimensional vectors is calculated. The larger s is, the greater the structural similarity between the scanned topology and the theoretical topology, and the greater the computational weight is assigned to the neighborhood priority sequence.

[0059] Optionally, in this embodiment, the first and second preset feature values ​​are obtained by selecting the first 7 based on the sorting results. Users can set them according to their actual situation, and there are no numerical restrictions here. This is just a preferred embodiment.

[0060] For each retained candidate matching point, the similarity of the corresponding descriptors is denoted as x1, and the similarity of the neighborhood priority sequence is denoted as x2. (1-s) is used as the weight of x1, and s is used as the weight of x2. The matching degree between the first matching point and each retained candidate matching point is obtained by weighted summation of x1 and x2. The retained candidate matching point with the highest matching degree is then used as the matching point of the first matching point in another image.

[0061] For the second matching point, the third matching point, and so on, their respective matching points are obtained using the same method. This results in multiple matching pairs between the depth map and the view map from the same perspective.

[0062] Furthermore, for each optimal matching point in each matching pair, the corresponding 3D point in the original 3D data (theoretical 3D model) is found by using the projection mapping relationship when generating the depth map, thus obtaining the theoretical point.

[0063] S104. Determine the original point corresponding to each vertex in the scan topology graph.

[0064] In this embodiment, the projection mapping relationship when generating the viewpoint map is used to find the corresponding 3D points in the 3D scanning model obtained by laser scanning, and the original points are obtained.

[0065] S105. Form a three-dimensional point correspondence sequence by combining the original points and theoretical points corresponding to the vertices in each scanned topology graph, and obtain the mapping relationship corresponding to the three-dimensional point correspondence sequence through the SIFT method.

[0066] In this embodiment, the original points and theoretical points corresponding to each vertex in the scanned topology map are combined to form a three-dimensional point correspondence sequence, resulting in a set of three-dimensional point correspondence sequences: {(3D point A1 in the three-dimensional scanning model obtained by laser scanning, 3D point B1 in the theoretical three-dimensional model), (A2, B2)...(An, Bn)}. Furthermore, the mapping relationship corresponding to the three-dimensional point correspondence sequence is obtained using the SIFT method.

[0067] S106. Based on the mapping relationship, the 3D scanning model, and the theoretical 3D model, obtain the fine registration transformation matrix.

[0068] In this embodiment, the fine registration transformation matrix is ​​obtained based on the mapping relationship, the 3D scanning model, and the theoretical 3D model, specifically including: The mapping relationship, the preset minimum point set size, the interior point distance threshold, and the maximum number of iterations are used as inputs to the RANSAC robust estimation algorithm, which outputs the rigid body transformation matrix. The rigid body transformation matrix, the 3D scanning model, and the theoretical 3D model are input into the ICP algorithm, and the ICP algorithm outputs the fine registration transformation matrix.

[0069] For example, due to information loss in the 3D-to-2D projection mapping, similarity ambiguity of feature descriptors on smooth surfaces, and processing errors and noise interference between the real workpiece and the ideal model, some incorrect matches may still exist in the corresponding point pairs obtained in the 3D space. Therefore, this embodiment uses a robust estimation algorithm (such as RANSAC) to solve for the optimal rigid body transformation matrix.

[0070] The mapping relationship between the corresponding sequences of 3D points obtained through the SIFT method, the preset minimum point set size (usually 3), the interior point distance threshold, and the maximum number of iterations are used as inputs. The RANSAC (Random Sample Consensus) robust estimation algorithm is then used for iterative calculation. This algorithm randomly selects the minimum point set to calculate the transformation hypothesis and verifies all point pairs, ultimately obtaining the rigid body transformation matrix. This rigid body transformation matrix represents the optimal rotation and translation parameters required for coarse alignment of the scanned point cloud with the CAD model. The interior point set corresponding to the rigid body transformation matrix represents the high-confidence, correctly matched point pairs selected from the initial matching that conform to the transformation, providing a clean and reliable data foundation for subsequent ICP fine matching.

[0071] The rigid body transformation matrix, the 3D scanning model, and the theoretical 3D model are input into the ICP algorithm, and the ICP algorithm outputs the fine registration transformation matrix.

[0072] S107. Decompose the fine registration transformation matrix to obtain the change amount, convert the change amount into control commands, and locate and adjust the defective links in the hyperbolic plate manufacturing process.

[0073] For example, based on the "rigid body transformation matrix" obtained in the previous step, the changes in the six degrees of freedom of the workpiece that need to be adjusted are decomposed (translations ΔX, ΔY, ΔZ along the X, Y, and Z axes and rotations θx, θy, θz around the X, Y, and Z axes).

[0074] Taking into account the kinematic model of the actuator (such as a robot gripper), these variables are converted into motion control commands for each positioner or robot joint.

[0075] The control system drives the positioning actuator (such as the gripper at the robot's end effector) to fine-tune the workpiece according to the calculated adjustment amount. After adjustment, the 3D scanner is restarted for re-measurement to acquire new point cloud data. The new point cloud is then quickly matched with the CAD model again to verify whether the current pose meets the tolerance requirements.

[0076] If the result is satisfactory: the process ends, the tooling is locked, and subsequent processing (such as drilling and milling) is performed. If the result is unsatisfactory: the fine registration transformation matrix, new point cloud data, and theoretical 3D model are input into the ICP algorithm. The ICP algorithm outputs a new fine registration transformation matrix, and the control commands are recalculated based on the new fine registration transformation matrix. Point cloud data is then acquired again, and this process is repeated for secondary fine-tuning, forming a closed-loop feedback system until the accuracy requirements are met. Typically, 1-2 iterations are sufficient to achieve extremely high accuracy.

[0077] This invention is now complete.

[0078] In summary, in this embodiment of the invention, a depth map-based SIFT matching method is used to first generate multi-view depth maps from the hyperbolic slab scan point cloud and CAD model, and extract robust SIFT features that are invariant to rotation and scaling. Subsequently, through high-dimensional descriptor matching and robust RANSAC estimation, a high-precision initial rigid body transformation matrix is ​​calculated, thereby achieving global coarse registration between the scan data and the design model. This effectively overcomes the matching ambiguity problem caused by the smooth surface and sparse features of the hyperbolic slab, providing a near-ideal iterative starting point for subsequent ICP matching. This allows it to focus on small local deformation optimization without consuming iterative resources in incorrect directions, thus significantly improving the accuracy, convergence speed, and overall robustness of the final ICP matching.

[0079] This invention also proposes an intelligent positioning system for hyperboloid plate processing; please refer to [link / reference]. Figure 4 The diagram shows a structural diagram of an intelligent positioning system for hyperboloid plate processing provided by an embodiment of the present invention. The system includes: a data acquisition module 101, a data processing module 102, and a defect location module 103.

[0080] The data acquisition module 101 is used to acquire the three-dimensional scanning model and theoretical three-dimensional model of the hyperbolic plate, and to acquire the scanning topology map and theoretical topology map based on the three-dimensional scanning model and theoretical three-dimensional model; The data processing module 102 is used to obtain the topological relationship strength of each vertex in the scanned topological graph and the topological relationship strength of each vertex in the theoretical topological graph, and sort them respectively to obtain the scan priority sort and the theoretical priority sort. Based on the scan priority sort and the theoretical priority sort, the optimal matching point of each vertex in the scanned topological graph in the theoretical topological graph is obtained, and the theoretical point corresponding to each optimal matching point is determined. The original point corresponding to each vertex in the scanned topological graph is determined. The original point and the theoretical point corresponding to each vertex in the scanned topological graph are combined to form a three-dimensional point correspondence sequence, and the mapping relationship corresponding to the three-dimensional point correspondence sequence is obtained by using the SIFT method. Based on the mapping relationship, the three-dimensional scan model and the theoretical three-dimensional model, the fine registration transformation matrix is ​​obtained. The defect location module 103 is used to decompose the fine registration transformation matrix to obtain the change amount, and convert the change amount into control commands to locate and adjust the defect links in the hyperbolic plate manufacturing process.

[0081] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the intelligent positioning system for hyperboloid plate processing and the intelligent positioning method for hyperboloid plate processing provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiment, which will not be repeated here.

[0082] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0083] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart positioning method for processing hyperboloid plates, characterized in that, include: Obtain the 3D scanning model and theoretical 3D model of the hyperbolic plate, and obtain the scanning topology diagram and theoretical topology diagram based on the 3D scanning model and theoretical 3D model; Obtain the topological relationship strength of each vertex in the scanned topological graph and the topological relationship strength of each vertex in the theoretical topological graph, and sort them respectively to obtain the scan priority sort and the theoretical priority sort. Based on the scanning priority sorting and theoretical priority sorting, obtain the optimal matching point of each vertex in the scanning topology graph in the theoretical topology graph, and determine the theoretical point corresponding to each optimal matching point; Determine the original point corresponding to each vertex in the scanned topology graph; The original points and theoretical points corresponding to each vertex in the scanned topology graph are combined into a 3D point correspondence sequence, and the mapping relationship corresponding to the 3D point correspondence sequence is obtained by using the SIFT method. Based on the mapping relationship, the 3D scanning model, and the theoretical 3D model, obtain the fine registration transformation matrix; The fine registration transformation matrix is ​​decomposed to obtain the change quantity, which is then converted into control commands to locate and adjust the defective links in the hyperbolic plate manufacturing process.

2. The intelligent positioning method for hyperboloid plate processing according to claim 1, characterized in that, The process of obtaining the scan topology map and theoretical topology map based on the 3D scan model and theoretical 3D model specifically includes: Viewpoint and depth maps are obtained from the 3D scanning model and the theoretical 3D model, respectively, with the same parameters for obtaining the viewpoint and depth maps. The SIFT algorithm is used to obtain the scan key points and theoretical key points corresponding to the view map and depth map, respectively. Using the Delaunay triangulation method, triangulation networks are constructed with scanned key points and theoretical key points as vertices, respectively, to obtain scanned topology and theoretical topology.

3. The intelligent positioning method for hyperboloid plate processing according to claim 1, characterized in that, The acquisition of the topological relation strength of each vertex in the scanned topological graph and the topological relation strength of each vertex in the theoretical topological graph specifically includes: Obtain the degree of each vertex in the topological graph, and determine the relative degree centrality of the Nth vertex by the ratio of the degree of the Nth vertex to the maximum degree of the vertex. The topological graph is either a scanned topological graph or a theoretical topological graph. The distances between the Nth vertex and other vertices in the topological graph are obtained by using Dijkstra's shortest path algorithm and summed to obtain the distance sum of the Nth vertex. The negative number of the distance sum of the Nth vertex is then used as the exponent of the natural base to obtain the centrality of the Nth vertex. The product of the relative degree centrality of the Nth vertex and the centrality of the Nth vertex is determined as the topological relation strength of the Nth vertex; Determine the topological relation strength of each vertex in the scanned topological graph and the topological relation strength of each vertex in the theoretical topological graph, respectively.

4. The intelligent positioning method for hyperboloid plate processing according to claim 1, characterized in that, The step of sorting by scan priority and theoretical priority to obtain the optimal matching point of each vertex in the scanned topology graph in the theoretical topology graph specifically includes: The vertex corresponding to the first element in the scan priority sorting in the scan topology graph is determined as the first matching point; Based on the first matching point, candidate matching points corresponding to the first matching point are determined from the theoretical priority ranking, wherein the absolute value of the difference between the degree of the first matching point and the degree of the candidate matching point is less than a preset degree threshold. Calculate the similarity between the descriptor of each candidate matching point and the descriptor of the first matching point, and determine the candidate matching points with similarity greater than a preset similarity threshold as retained candidate matching points; Determine the priority sequence of neighboring points of the first matching point based on the scanned topology map; Determine the neighborhood priority sequence of candidate matching points to be retained based on the theoretical topology graph; Obtain the cosine similarity between the scanned topology graph and the theoretical topology graph; The matching degree between the first matching point and each retained candidate matching point is calculated based on the cosine similarity. The retained candidate matching point with the highest matching degree is determined as the optimal matching point of the first matching point. The vertex corresponding to the Ath element in the scan priority sorting in the scan topology graph is determined as the Ath matching point; Obtain the optimal matching point of the vertex corresponding to each element in the scan topology graph in the scan priority sorting.

5. The intelligent positioning method for hyperboloid plate processing according to claim 4, characterized in that, The step of determining the priority sequence of neighboring points of the first matching point based on the scanned topology map specifically includes: The neighborhood points of the first matching point are determined from the scanned topology graph, wherein the neighborhood points of the first matching point and the first matching point are connected in the scanned topology graph through the edges of the scanned topology graph; Calculate the similarity between the descriptors of the neighboring points of each first matching point and the descriptor of the first matching point, and sort them in ascending order to obtain the sequence of neighboring points of the first matching point; The index value of each element in the scan priority sort is determined as the priority sort value of the corresponding vertex in the scan topology graph, thus obtaining the priority sort value of each vertex in the scan topology graph; The value of each element in the neighborhood point sequence of the first matching point is replaced with the priority sorting value of the corresponding vertex in the scanned topology graph, to obtain the neighborhood point priority sequence.

6. The intelligent positioning method for hyperboloid plate processing according to claim 4, characterized in that, The step of determining the neighborhood priority sequence of candidate matching points based on the theoretical topology graph specifically includes: The neighborhood points of the candidate matching points are determined from the theoretical topology graph, wherein the neighborhood points of the candidate matching points and the candidate matching points are connected by edges in the theoretical topology graph. Calculate the similarity between the descriptors of the neighboring points of each retained candidate matching point and the descriptors of the retained candidate matching points, and sort them in ascending order to obtain the sequence of neighboring points of the retained candidate matching points; The index value of each element in the theoretical priority sort is determined as the priority sort value of the corresponding vertex in the theoretical topology graph, thus obtaining the priority sort value of each vertex in the theoretical topology graph. The value of each element in the neighborhood point sequence of the retained candidate matching point is replaced with the priority ranking value of the corresponding vertex in the theoretical topology graph, thus obtaining the neighborhood priority sequence.

7. The intelligent positioning method for hyperboloid plate processing according to claim 4, characterized in that, The acquisition of the cosine similarity between the scanned topology graph and the theoretical topology graph specifically includes: Calculate the eigenvalues ​​of the normalized Laplacian matrix of the scanned topology graph, and sort the eigenvalues ​​in descending order to obtain the first eigenvalue sequence; Calculate the eigenvalues ​​of the normalized Laplacian matrix of the theoretical topology graph, and arrange the eigenvalues ​​in descending order to obtain the second eigenvalue sequence; Obtain the elements in the first feature value sequence that are greater than the preset first feature value to obtain the first vector; Obtain the elements in the second feature value sequence that are greater than the preset second feature value to obtain the second vector; Calculate the cosine similarity between the first vector and the second vector, and determine it as the cosine similarity between the scanned topological graph and the theoretical topological graph.

8. The intelligent positioning method for hyperboloid plate processing according to claim 4, characterized in that, The step of calculating the matching degree between the first matching point and each retained candidate matching point based on cosine similarity specifically includes: The similarity between the descriptor of the Mth retained candidate matching point and the descriptor of the first matching point is determined as the similarity of the Mth descriptor; The cosine similarity is negative, then 1 is added, and multiplied with the similarity of the Mth descriptor to obtain the first matching degree of the Mth retained candidate matching point; Calculate the similarity between the priority sequence of neighboring points and the priority sequence of the neighboring points of the Mth retained candidate matching point to obtain the similarity of the Mth neighborhood. The product of the similarity of the Mth neighborhood and the cosine similarity is determined as the second matching degree of the Mth retained candidate matching point; The product of the first matching degree and the second matching degree of the Mth retained candidate matching point is determined as the matching degree between the first matching point and the Mth retained candidate matching point. Obtain the matching degree between the first matching point and each retained candidate matching point.

9. The intelligent positioning method for hyperboloid plate processing according to claim 4, characterized in that, The process of obtaining the fine registration transformation matrix based on the mapping relationship, the 3D scanning model, and the theoretical 3D model specifically includes: The mapping relationship, the preset minimum point set size, the interior point distance threshold, and the maximum number of iterations are used as inputs to the RANSAC robust estimation algorithm, which outputs the rigid body transformation matrix. The rigid body transformation matrix, the 3D scanning model, and the theoretical 3D model are input into the ICP algorithm, and the ICP algorithm outputs the fine registration transformation matrix.

10. An intelligent positioning system for processing hyperboloid plates, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent positioning method for hyperboloid plate processing as described in any one of claims 1-9.

Images