Three-dimensional point cloud registration method, system and equipment based on graph voting and storage medium

Abstract

The invention relates to the technical field of point cloud registration, in particular to a three-dimensional point cloud registration method, system and device based on graph voting and a storage medium, and aims to solve the technical problem of low point cloud registration precision in the prior art. Obtaining vertexes and edges in a neighborhood range of each vertex, and obtaining a vertex sub-graph of each vertex; then, voting and screening are carried out based on the vertex repetition times among the vertex subgraphs, and the point cloud pairs corresponding to the vertexes with the voting values larger than the voting value threshold value serve as selected point cloud pairs; and finally, point cloud pairs are selected from the selected point cloud pair set to calculate an optimal spatial variation matrix, point cloud registration of the source point cloud set and the target point cloud set is carried out, the accuracy and precision of point cloud registration are improved, the calculation complexity is reduced, the registration efficiency is improved, and the method is suitable for efficient and accurate registration of large-scale point cloud data.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud registration, and specifically to a three-dimensional point cloud registration method, system, device, and storage medium based on graph voting. Background Technique

[0002] Three-dimensional point cloud registration technology is widely used in fields such as SLAM (Simultaneous Localization and Mapping) and three-dimensional reconstruction. Existing mainstream methods (such as the Teaser method, SC2-PCR method, and MAC method) rely on the quality of the initial correspondence relationship and have the following problems: high computational complexity: existing methods need to traverse all possible point pair relationships. When the number of point pairs is relatively large, a large amount of complex information needs to be processed, which is not only time-consuming but also requires a large amount of memory, making it difficult to process large-scale point clouds; sensitive to noise: incorrect point pairs in the initial point pair data will significantly reduce the registration accuracy. Summary of the Invention

[0003] The purpose of the present invention is to provide a three-dimensional point cloud registration method, system, device, and storage medium based on graph voting.

[0004] The technical solution of the present invention is as follows: A three-dimensional point cloud registration method based on graph voting includes the following operations: S1. Based on the source point cloud set and the target point cloud set, obtain a number of point cloud pairs to form a point cloud pair set; use the point cloud pairs as vertices, and connect two vertices with a vertex distance less than the vertex distance threshold with an edge to obtain a first-order graph; from the first-order graph, obtain the vertices and edges within the neighborhood range of each vertex to obtain the vertex subgraph of each vertex; S2. Based on the vertex repetition times between different vertex subgraphs, perform vertex voting to obtain the voting value of each vertex; use the point cloud pairs corresponding to the vertices with a voting value greater than the voting value threshold as the selected point cloud pairs to form a selected point cloud pair set; S3. Randomly select a number of selected point cloud pairs from the selected point cloud pair set, calculate the point cloud spatial transformation matrix, perform a spatial transformation on the target point cloud set to obtain a spatially transformed point cloud set; obtain the point cloud distance of each point cloud pair between the spatially transformed point cloud set and the source point cloud set, and use the point cloud pairs with a point cloud distance less than the point cloud distance threshold as the feature point cloud pairs to obtain the number of feature point cloud pairs; if the number of feature point cloud pairs is less than the point cloud pair number threshold, record the point cloud spatial transformation matrix; repeat the operations of selecting the selected point cloud pairs, calculating the point cloud spatial transformation matrix, obtaining the number of feature point cloud pairs, and recording the point cloud spatial transformation matrix several times until the number of repetitions is equal to the repetition number threshold, and use the point cloud spatial transformation matrix corresponding to the minimum value of the number of feature point cloud pairs as the optimal point cloud spatial transformation matrix; based on the optimal point cloud spatial transformation matrix, perform point cloud registration on the source point cloud set and the target point cloud set.

[0005] The vertex distance in S1 is calculated by the following formula: , is the vertex and the vertex the vertex distance between them, is the Euclidean distance between the source point cloud and the source point cloud in the source point cloud set, is the Euclidean distance between the target point cloud and the target point cloud in the target point cloud set, , .

[0006] The operation to obtain the voting value of each vertex in S2 is: obtain the total number of times each vertex appears repeatedly between two different vertex subgraphs as the voting value of each vertex.

[0007] The operation to obtain the voting value of each vertex in S2 is: take the vertex subgraphs corresponding to the vertices existing in the vertex subgraph of the current vertex as the vertex subgraphs to be compared with the current vertex, and the vertex subgraph of the current vertex as the target vertex subgraph; obtain the total number of times each vertex appears repeatedly between all target vertex subgraphs and the corresponding vertex subgraphs to be compared with the current vertex as the voting value of each vertex.

[0008] The method to obtain the point cloud spatial transformation matrix in S3 is: obtain the point clouds belonging to the target point cloud set and the source point cloud set respectively from several selected point cloud pairs to obtain several target selected point clouds and several source selected point clouds; take the average values of all target selected point clouds and all source selected point clouds as the target selected point cloud center and the source selected point cloud center respectively; subtract the corresponding target selected point cloud center and source selected point cloud center from each target selected point cloud and each source selected point cloud to obtain several target de-centered point clouds and several source de-centered point clouds; construct a point cloud covariance matrix based on the several target de-centered point clouds and several source de-centered point clouds; perform singular value decomposition on the point cloud covariance matrix to obtain a rotation matrix; obtain a translation vector based on the rotation matrix; obtain the point cloud spatial transformation matrix based on the rotation matrix and the translation vector.

[0009] The point cloud spatial transformation matrix is , is the translation vector, is the rotation matrix.

[0010] In S1, the vertices and edges within the neighborhood range of each vertex are the vertices and edges directly connected to each vertex.

[0011] A 3D point cloud registration system based on graph voting, which is used to implement the above-mentioned 3D point cloud registration method based on graph voting, includes: A vertex subgraph generation module, which is used to obtain a number of point cloud pairs based on a source point cloud set and a target point cloud set to form a point cloud pair set; use the point cloud pairs as vertices, connect two vertices with a vertex distance less than the vertex distance threshold with an edge to obtain a first-order graph; from the first-order graph, obtain the vertices and edges within the neighborhood range of each vertex to obtain the vertex subgraph of each vertex; A selected point cloud pair set generation module, which is used to perform vertex voting based on the vertex repetition times between different vertex subgraphs to obtain the voting value of each vertex; use the point cloud pairs corresponding to the vertices with a voting value greater than the voting value threshold as selected point cloud pairs to form a selected point cloud pair set; A point cloud registration module, which is used to randomly select several selected point cloud pairs from the selected point cloud pair set, calculate the point cloud space transformation matrix, perform a spatial transformation on the target point cloud set to obtain a spatially transformed point cloud set; obtain the point cloud distance of each point cloud pair between the spatially transformed point cloud set and the source point cloud set, use the point cloud pairs with a point cloud distance less than the point cloud distance threshold as feature point cloud pairs, and obtain the number of feature point cloud pairs; if the number of feature point cloud pairs is less than the point cloud pair number threshold, record the point cloud space transformation matrix; repeat the operations of selecting selected point cloud pairs, calculating the point cloud space transformation matrix, obtaining the number of feature point cloud pairs, and recording the point cloud space transformation matrix several times until the number of repetitions is equal to the repetition number threshold, and use the point cloud space transformation matrix corresponding to the minimum value of the number of feature point cloud pairs as the optimal point cloud space transformation matrix; based on the optimal point cloud space transformation matrix, perform point cloud registration of the source point cloud set and the target point cloud set.

[0012] A 3D point cloud registration device based on graph voting, including a processor and a memory. Among them, when the processor executes the computer program stored in the memory, it implements the above-mentioned 3D point cloud registration method based on graph voting.

[0013] A computer-readable storage medium, which is characterized in that it is used to store a computer program. Among them, when the computer program is executed by a processor, it implements the above-mentioned 3D point cloud registration method based on graph voting.

[0014] The beneficial effects of the present invention are as follows: A 3D point cloud registration method based on graph voting provided by the present invention. First, taking point cloud pairs as vertices, a first-order graph of point cloud pairs is constructed based on the vertex distance, and vertices and edges within the neighborhood range of each vertex are obtained. Noise point pairs are filtered to obtain the vertex subgraph of each vertex. Then, voting and screening are performed based on the number of repetitions of vertices between vertex subgraphs. The point cloud pairs corresponding to the vertices with voting values greater than the voting value threshold are used as selected point cloud pairs, and high-quality matching point cloud pairs that are recognized in multiple vertex subgraphs are retained, making the selected point cloud pair set more representative and reliable. Finally, the spatial transformation matrix is calculated by selecting point cloud pairs from the selected point cloud pair set. After the target point cloud is transformed, feature point cloud pairs are screened and the quantity is counted. After multiple iterations, the matrix with the minimum number of feature point cloud pairs is selected as the optimal point cloud spatial transformation matrix for point cloud registration of the source point cloud set and the target point cloud set, improving the accuracy and precision of point cloud registration, reducing the computational complexity, enhancing the registration efficiency, and being suitable for efficient and accurate registration of large-scale point cloud data. Detailed implementation mode

[0015] This embodiment provides a 3D point cloud registration method based on graph voting, including the following operations: S1. Based on the source point cloud set and the target point cloud set, a number of point cloud pairs are obtained to form a point cloud pair set. Taking the point cloud pairs as vertices, two vertices with a vertex distance less than the vertex distance threshold are connected by an edge to obtain a first-order graph. From the first-order graph, vertices and edges within the neighborhood range of each vertex are obtained to obtain the vertex subgraph of each vertex. S2. Based on the number of repetitions of vertices between different vertex subgraphs, vertex voting is performed to obtain the voting value of each vertex. The point cloud pairs corresponding to the vertices with voting values greater than the voting value threshold are used as selected point cloud pairs to form a selected point cloud pair set. S3. Randomly select a number of selected point cloud pairs from the selected point cloud pair set, calculate the point cloud spatial transformation matrix, perform spatial transformation on the target point cloud set to obtain a spatially transformed point cloud set. Obtain the point cloud distance of each point cloud pair between the spatially transformed point cloud set and the source point cloud set. The point cloud pairs with a point cloud distance less than the point cloud distance threshold are used as feature point cloud pairs, and the number of feature point cloud pairs is obtained. If the number of feature point cloud pairs is less than the point cloud pair number threshold, record the point cloud spatial transformation matrix. Repeat the operations of selecting selected point cloud pairs, calculating the point cloud spatial transformation matrix, obtaining the number of feature point cloud pairs, and recording the point cloud spatial transformation matrix several times until the number of repetitions is equal to the repetition number threshold. The point cloud spatial transformation matrix corresponding to the minimum value of the number of feature point cloud pairs is used as the optimal point cloud spatial transformation matrix. Based on the optimal point cloud spatial transformation matrix, point cloud registration of the source point cloud set and the target point cloud set is performed.

[0016] The specific operation details are as follows.

[0017] S1. Based on the source point cloud set and the target point cloud set, a number of point cloud pairs are obtained to form a point cloud pair set; taking the point cloud pairs as vertices, two vertices with a vertex distance less than the vertex distance threshold are connected by an edge to obtain a first-order graph; from the first-order graph, the vertices and edges directly connected to each vertex are obtained to get the vertex subgraph of each vertex.

[0018] Construct a first-order graph of point cloud pairs based on the vertex distance and extract the vertex subgraph, filter out the noise point pairs, and use the local geometric consistency constraint of the subgraph to retain the point cloud pairs with similar structures, providing a low-noise and highly correlated sample set for subsequent point cloud registration, reducing redundant calculations, and improving the registration efficiency and accuracy.

[0019] First, based on the source point cloud set and the target point cloud set, using manual marking or feature-based methods, a number of point cloud pairs are obtained to form a point cloud pair set.

[0020] Then, taking the point cloud pairs as vertices, two vertices with a vertex distance less than the vertex distance threshold are connected by an edge to obtain a first-order graph that constrains the rigid distance between the pair of vertices (point cloud pairs).

[0021] The vertex distance is calculated by the following formula: , is the vertex distance between vertex (point cloud pair) and vertex (point cloud pair) , is the Euclidean distance between source point cloud and source point cloud in the source point cloud set, is the Euclidean distance between target point cloud and target point cloud in the target point cloud set, , .

[0022] Finally, from the first-order graph, the vertices and edges within the neighborhood range of each vertex are obtained. These vertices and edges usually have consistent transformation parameters with the central vertex. Preferably, the vertices and edges directly connected to each vertex are obtained to get the vertex subgraph of each vertex. The vertex subgraph reflects all compatible local consistencies of the nodes and is used to prevent redundant samples (such as highly similar repeated matching pairs) from entering the subsequent point cloud registration process, reducing the amount of calculation.

[0023] S2. Based on the number of vertex repetitions between different vertex subgraphs, vertex voting is performed to obtain the voting value of each vertex; the point cloud pairs corresponding to the vertices with voting values greater than the voting value threshold are used as the selected point cloud pairs to form a selected point cloud pair set.

[0024] Voting and screening based on the number of vertex repetitions between vertex subgraphs can utilize the global structural consistency to identify and eliminate local noisy point cloud pairs, retain high-quality matching point cloud pairs that are recognized in multiple vertex subgraphs, make the selected point cloud pair set more representative and reliable, provide high-quality samples for subsequent point cloud registration, greatly reduce invalid calculations, accelerate convergence, and improve the registration accuracy and robustness.

[0025] First, based on the number of vertex repetitions between different vertex subgraphs, vertex voting is performed to obtain the voting value of each vertex.

[0026] The operation of obtaining the voting value of each vertex is as follows: Obtain the total number of times each vertex appears repeatedly between two different vertex subgraphs as the voting value of each vertex.

[0027] To further reduce the computational amount and improve the computational efficiency, the operation of obtaining the voting value of each vertex can also be: Take the vertex subgraphs corresponding to the vertices existing in the vertex subgraph of the current vertex as the vertex subgraphs to be compared with the current vertex, and the vertex subgraph of the current vertex as the target vertex subgraph; Obtain the total number of times each vertex appears repeatedly between all target vertex subgraphs and the corresponding vertex subgraphs to be compared as the voting value of each vertex.

[0028] Next, the point cloud pairs corresponding to the vertices with voting values greater than the voting value threshold are used as the selected point cloud pairs, and all the selected point cloud pairs form a selected point cloud pair set.

[0029] S3. Based on the selected point cloud pair set, obtain the optimal point cloud spatial transformation matrix, and based on the optimal point cloud spatial transformation matrix, perform point cloud registration on the source point cloud set and the target point cloud set.

[0030] By calculating the spatial transformation matrix from the point cloud pairs selected from the selected point cloud pair set, screening the feature point cloud pairs and counting the number after transforming the target point cloud, and selecting the matrix with the smallest number of feature point cloud pairs after multiple iterations as the optimal point cloud spatial transformation matrix, it can effectively exclude the mis-matched point pairs in the point cloud pair set, reduce the interference of outliers, and thus improve the accuracy of point cloud registration; At the same time, through threshold filtering and iterative optimization, it can avoid falling into local optimality, and only needs to process the selected point cloud pairs instead of all data, reducing the computational complexity and improving the registration efficiency; Moreover, based on the characteristics of sampling and statistics, it can adapt to the challenges of noise and redundant points in large-scale point cloud data, avoid the resource consumption bottleneck of all-data calculation, and is suitable for the efficient and accurate registration of large-scale point cloud data.

[0031] The operation steps of obtaining the optimal point cloud spatial transformation matrix based on the selected point cloud pair set are as follows.

[0032] Step 1: Randomly select several pairs of selected point clouds from the set of selected point cloud pairs, and calculate the point cloud spatial transformation matrix; based on the point cloud spatial transformation matrix, perform a spatial transformation on the target point cloud set to obtain a spatially transformed point cloud set.

[0033] The steps to obtain the point cloud spatial transformation matrix are as follows.

[0034] Step a: Obtain the point clouds in several pairs of selected point clouds that belong to the target point cloud set and the source point cloud set respectively, to obtain several target selected point clouds and several source selected point clouds; take the average value of all target selected point clouds and the average value of all source selected point clouds as the target selected point cloud center and the source selected point cloud center respectively.

[0035] Step b: Subtract the corresponding target selected point cloud center and source selected point cloud center from each target selected point cloud and each source selected point cloud to obtain several target de-centered point clouds and several source de-centered point clouds; based on the several target de-centered point clouds and several source de-centered point clouds, construct a point cloud covariance matrix.

[0036] The point cloud covariance matrix is calculated by the following formula: , is the point cloud covariance matrix, , are respectively the i th abscissa of the target de-centered point cloud, the i th abscissa of the source de-centered point cloud, , are respectively the i th ordinate of the target de-centered point cloud, the i th ordinate of the source de-centered point cloud, , are respectively the i th vertical coordinate of the target de-centered point cloud, the i th vertical coordinate of the source de-centered point cloud.

[0037] Step c: Perform a singular value decomposition on the point cloud covariance matrix to obtain a rotation matrix; based on the rotation matrix, obtain a translation vector; based on the rotation matrix and the translation vector, obtain the point cloud spatial transformation matrix.

[0038] The above-mentioned obtaining of the translation vector based on the rotation matrix is calculated by the following formula: , is the translation vector, is the rotation matrix, , are respectively the target selected point cloud center and the source selected point cloud center.

[0039] The above point cloud spatial transformation matrix is .

[0040] Step 2: Obtain the point cloud distance between each point cloud pair of the spatially transformed point cloud set and the source point cloud set. Consider the point cloud pairs with a point cloud distance (the Euclidean distance between two point clouds in a point cloud pair) less than the point cloud distance threshold as feature point cloud pairs, and obtain the number of feature point cloud pairs. If the number of feature point cloud pairs is less than the point cloud pair number threshold, record the point cloud spatial transformation matrix.

[0041] Step 3: Repeat the operations of selecting high-quality point cloud pairs, calculating the point cloud spatial transformation matrix, obtaining the number of feature point cloud pairs, and recording the point cloud spatial transformation matrix several times until the number of repetitions is equal to the repetition number threshold. Take the point cloud spatial transformation matrix corresponding to the minimum value of the number of feature point cloud pairs as the optimal point cloud spatial transformation matrix.

[0042] Finally, based on the optimal point cloud spatial transformation matrix, perform point cloud registration on the source point cloud set and the target point cloud set. Specifically, it is obtained by multiplying the optimal point cloud spatial transformation matrix with the homogeneous coordinate matrix corresponding to the target point cloud set.

[0043] This embodiment also provides a three-dimensional point cloud registration system based on graph voting for implementing the above three-dimensional point cloud registration method based on graph voting, including: A vertex subgraph generation module, which is used to obtain a number of point cloud pairs based on the source point cloud set and the target point cloud set to form a point cloud pair set; use the point cloud pairs as vertices, connect two vertices with a vertex distance less than the vertex distance threshold with an edge to obtain a first-order graph; from the first-order graph, obtain the vertices and edges within the neighborhood range of each vertex to obtain the vertex subgraph of each vertex; A high-quality point cloud pair set generation module, which is used to perform vertex voting based on the vertex repetition times between different vertex subgraphs to obtain the voting value of each vertex; consider the point cloud pairs corresponding to the vertices with a voting value greater than the voting value threshold as high-quality point cloud pairs to form a high-quality point cloud pair set; A point cloud registration module, which is used to randomly select several high-quality point cloud pairs from the high-quality point cloud pair set, calculate the point cloud spatial transformation matrix, perform spatial transformation on the target point cloud set to obtain a spatially transformed point cloud set; obtain the point cloud distance between each point cloud pair of the spatially transformed point cloud set and the source point cloud set, consider the point cloud pairs with a point cloud distance less than the point cloud distance threshold as feature point cloud pairs, and obtain the number of feature point cloud pairs; if the number of feature point cloud pairs is less than the point cloud pair number threshold, record the point cloud spatial transformation matrix; repeat the operations of selecting high-quality point cloud pairs, calculating the point cloud spatial transformation matrix, obtaining the number of feature point cloud pairs, and recording the point cloud spatial transformation matrix several times until the number of repetitions is equal to the repetition number threshold. Take the point cloud spatial transformation matrix corresponding to the minimum value of the number of feature point cloud pairs as the optimal point cloud spatial transformation matrix; based on the optimal point cloud spatial transformation matrix, perform point cloud registration on the source point cloud set and the target point cloud set.

[0044] This embodiment also provides a three-dimensional point cloud registration device based on graph voting, including a processor and a memory. When the processor executes the computer program stored in the memory, the above-mentioned three-dimensional point cloud registration method based on graph voting is implemented.

[0045] This embodiment also provides a computer-readable storage medium, which is characterized in that it is used to store a computer program. When the computer program is executed by a processor, the above-mentioned three-dimensional point cloud registration method based on graph voting is implemented.

[0046] A three-dimensional point cloud registration method based on graph voting provided in this embodiment. First, the point cloud pair is used as a vertex, and a first-order graph of the point cloud pair is constructed based on the vertex distance, and the vertices and edges within the neighborhood range of each vertex are obtained, and the noise point pairs are filtered to obtain the vertex subgraph of each vertex; then, voting screening is performed based on the number of repeated vertices between vertex subgraphs, and the point cloud pair corresponding to the vertex with a voting value greater than the voting value threshold is used as the selected point cloud pair, and the high-quality matching point cloud pairs that are recognized in multiple vertex subgraphs are retained, making the selected point cloud pair set more representative and reliable; finally, the spatial transformation matrix is calculated by selecting point cloud pairs from the selected point cloud pair set, the feature point cloud pairs are screened and the quantity is counted after the target point cloud is transformed, and the matrix with the smallest number of feature point cloud pairs is selected as the optimal point cloud spatial transformation matrix after multiple iterations, and the point cloud registration of the source point cloud set and the target point cloud set is performed, improving the accuracy and precision of point cloud registration, reducing the computational complexity, enhancing the registration efficiency, and being suitable for the efficient and accurate registration of large-scale point cloud data.

Claims

1. A 3D point cloud registration method based on graph voting, characterized in that Including the following operations: S1. Based on the source point cloud set and the target point cloud set, obtain a number of point cloud pairs to form a point cloud pair set; Use the point cloud pairs as vertices, and connect two vertices with a vertex distance less than the vertex distance threshold with an edge to obtain a first-order graph; From the first-order graph, obtain the vertices and edges within the neighborhood range of each vertex to obtain the vertex subgraph of each vertex; S2. Based on the number of times vertices are repeated between different vertex subgraphs, perform vertex voting to obtain the voting value of each vertex; use the point cloud pairs corresponding to the vertices with a voting value greater than the voting value threshold as the selected point cloud pairs to form a selected point cloud pair set; S3. Randomly select a number of selected point cloud pairs from the selected point cloud pair set, calculate the point cloud space transformation matrix, perform space transformation on the target point cloud set to obtain a space-transformed point cloud set; Obtain the point cloud distance between each point cloud pair of the space-transformed point cloud set and the source point cloud set, use the point cloud pairs with a point cloud distance less than the point cloud distance threshold as the feature point cloud pairs, and obtain the number of feature point cloud pairs; if the number of feature point cloud pairs is less than the point cloud pair number threshold, record the point cloud space transformation matrix; Repeat the operations of selecting selected point cloud pairs, calculating the point cloud space transformation matrix, obtaining the number of feature point cloud pairs, and recording the point cloud space transformation matrix several times until the number of repetitions is equal to the repetition number threshold, and use the point cloud space transformation matrix corresponding to the minimum value of the number of feature point cloud pairs as the optimal point cloud space transformation matrix; Based on the optimal point cloud space transformation matrix, perform point cloud registration of the source point cloud set and the target point cloud set.

2. The three-dimensional point cloud registration method based on graph voting according to claim 1, characterized in that The vertex distance in S1 is calculated by the following formula: ， is the vertex and the vertex The vertex distance between is the Euclidean distance between the source point cloud and the source point cloud in the source point cloud set, is the Euclidean distance between the target point cloud and the target point cloud in the target point cloud set, , .

3. The 3D point cloud registration method based on graph voting according to claim 1, wherein The operation to obtain the voting value of each vertex in S2 is: obtain the total number of times each vertex appears repeatedly between two different vertex subgraphs as the voting value of each vertex.

4. The 3D point cloud registration method based on graph voting according to claim 1, characterized in that The operation to obtain the voting value of each vertex in S2 is: use the vertex subgraphs corresponding to the vertices existing in the vertex subgraph of the current vertex as the vertex subgraphs to be compared with the current vertex, and use the vertex subgraph of the current vertex as the target vertex subgraph; obtain the total number of times each vertex appears repeatedly between all target vertex subgraphs and the corresponding vertex subgraphs to be compared as the voting value of each vertex.

5. The 3D point cloud registration method based on graph voting according to claim 1, wherein, The method for obtaining the point cloud space transformation matrix in S3 is: Obtain the point clouds belonging to the target point cloud set and the source point cloud set respectively from a number of selected point cloud pairs to obtain a number of target selected point clouds and a number of source selected point clouds; use the average value of all target selected point clouds and the average value of all source selected point clouds as the target selected point cloud center and the source selected point cloud center respectively; Subtract the corresponding target selected point cloud center and source selected point cloud center from each target selected point cloud and each source selected point cloud to obtain a number of target de-centered point clouds and a number of source de-centered point clouds; based on the number of target de-centered point clouds and the number of source de-centered point clouds, construct a point cloud covariance matrix; Perform singular value decomposition on the point cloud covariance matrix to obtain a rotation matrix; Based on the rotation matrix, obtain a translation vector; Based on the rotation matrix and the translation vector, obtain the point cloud space transformation matrix.

6. The three-dimensional point cloud registration method based on graph voting according to claim 5, characterized in that The point cloud spatial transformation matrix is , is the translation vector, is the rotation matrix.

7. The 3D point cloud registration method based on graph voting according to claim 1, characterized in that In S1, the vertices and edges within the neighborhood range of each vertex are the vertices and edges directly connected to each vertex.

8. A three-dimensional point cloud registration system based on graph voting, which is used to implement the three-dimensional point cloud registration method based on graph voting described in claim 1, characterized in that, It includes: A vertex subgraph generation module, which is used to obtain a number of point cloud pairs based on a source point cloud set and a target point cloud set, forming a point cloud pair set; Taking the point cloud pairs as vertices, connecting two vertices with a vertex distance less than the vertex distance threshold with an edge to obtain a first-order graph; from the first-order graph, obtaining the vertices and edges within the neighborhood range of each vertex to obtain the vertex subgraph of each vertex; A selected point cloud pair set generation module, which is used to perform vertex voting based on the vertex repetition times between different vertex subgraphs to obtain the voting value of each vertex; taking the point cloud pairs corresponding to the vertices with a voting value greater than the voting value threshold as selected point cloud pairs, forming a selected point cloud pair set; A point cloud registration module, which is used to randomly select several selected point cloud pairs from the selected point cloud pair set, calculate the point cloud spatial transformation matrix, perform spatial transformation on the target point cloud set to obtain a spatially transformed point cloud set; Obtaining the point cloud distance of each point cloud pair between the spatially transformed point cloud set and the source point cloud set, taking the point cloud pairs with a point cloud distance less than the point cloud distance threshold as feature point cloud pairs, and obtaining the number of feature point cloud pairs; if the number of feature point cloud pairs is less than the point cloud pair number threshold, recording the point cloud spatial transformation matrix; repeating the operations of selecting selected point cloud pairs, calculating the point cloud spatial transformation matrix, obtaining the number of feature point cloud pairs, and recording the point cloud spatial transformation matrix several times until the number of repetitions is equal to the repetition number threshold, taking the point cloud spatial transformation matrix corresponding to the minimum value of the number of feature point cloud pairs as the optimal point cloud spatial transformation matrix; based on the optimal point cloud spatial transformation matrix, performing point cloud registration on the source point cloud set and the target point cloud set.

9. A three-dimensional point cloud registration device based on graph voting, characterized in that, It includes a processor and a memory. Among them, when the processor executes the computer program stored in the memory, it implements the three-dimensional point cloud registration method based on graph voting as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It is used to store a computer program. Among them, when the computer program is executed by a processor, it implements the three-dimensional point cloud registration method based on graph voting as described in any one of claims 1-7.

Images