Point cloud registration method and device for automobile steering knuckle
By downsampling and feature point detection of complex mechanical parts point clouds, and precise registration combined with Euclidean distance weight and error function, the problem of low registration efficiency of complex mechanical parts point clouds in the existing technology is solved, achieving more efficient and precise registration.
Patent Information
- Application Number
- CN202411530438.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-30
- Publication Date
- 2025-02-14
AI Technical Summary
In the prior art, the registration efficiency of point clouds of complex mechanical parts is low, which affects the task completion efficiency and quality.
By obtaining the source point cloud set and the target point cloud set, voxel grid division and downsampling are performed, feature points are detected and registered, weights are set according to Euclidean distance, and precise registration is performed using a preset error function.
The speed and accuracy of point cloud registration are improved, and the problem of low registration efficiency in point cloud registration tasks for complex mechanical parts is solved.
Smart Images
Figure CN119444817A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of point cloud registration of an automobile steering knuckle, and in particular to a point cloud registration method and device for an automobile steering knuckle. Background Art
[0002] In industrial application scenarios such as 3D measurement of complex mechanical parts such as automobile steering knuckles and automated robot grasping, the speed of point cloud registration directly affects the efficiency of task completion, and the accuracy of point cloud registration determines the accuracy of pose estimation, thus affecting the quality of task completion. For the registration task of point clouds of complex mechanical parts, the registration efficiency is low.
[0003] There is currently no effective solution to the above-mentioned problems in the related art. Summary of the invention
[0004] The present application provides a point cloud registration method and device for an automobile steering knuckle, so as to solve the problem of low registration efficiency in the prior art for point cloud registration tasks of complex mechanical parts.
[0005] In a first aspect, the present application provides a point cloud registration method for an automobile steering knuckle, comprising: obtaining a source point cloud set and a target point cloud set corresponding to the steering knuckle, wherein the point clouds in the source point cloud set are point clouds obtained at different angles, and the point clouds in the target point cloud set are preset point clouds; dividing the point clouds in the source point cloud set into voxel grids, and downsampling the point clouds in the source point cloud set based on the distance between the points in the voxel grid and the centroids in the voxel grid; performing feature point detection on the downsampled source point cloud set and the target point cloud, and registering the downsampled source point cloud set and the target point cloud based on the feature points; determining a plurality of point cloud pairs from the downsampled source point cloud set and the target point cloud, setting different weights for point cloud pairs with different Euclidean distances, and obtaining a registered point cloud set based on a preset error function, wherein the point cloud pair refers to a point cloud pair in which the Euclidean distance between a first point cloud in the downsampled source point cloud set and a second point cloud in the target point cloud is less than a preset threshold.
[0006] In a second aspect, the present application provides a point cloud registration device for an automobile steering knuckle, comprising: an acquisition module, used to acquire a source point cloud set and a target point cloud set corresponding to the steering knuckle, wherein the point clouds in the source point cloud set are point clouds acquired at different angles, and the point clouds in the target point cloud set are preset point clouds; a first processing module, used to divide the point clouds in the source point cloud set into voxel grids, and downsample the point clouds in the source point cloud set based on the distance between the points in the voxel grid and the centroid in the voxel grid; a second processing module, used to downsample the point clouds in the source point cloud set; The downsampled source point cloud set and the target point cloud are subjected to feature point detection, and the downsampled source point cloud set and the target point cloud are registered based on the feature points; a third processing module is used to determine a plurality of point cloud pairs from the downsampled source point cloud set and the target point cloud, set different weights for point cloud pairs with different Euclidean distances, and obtain the registered point cloud set based on a preset error function, wherein the point cloud pair refers to a point cloud pair whose Euclidean distance between the first point cloud of the downsampled source point cloud set and the second point cloud in the target point cloud is less than a preset threshold.
[0007] In a third aspect, the present application provides an electronic device comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; and at least one memory connected to the at least one bus, wherein the processor is configured to execute the point cloud registration method for the automobile steering knuckle described in the first aspect of the present application.
[0008] In a fourth aspect, the present application further provides a computer storage medium storing computer executable instructions, wherein the computer executable instructions are used to execute the point cloud registration method for the automobile steering knuckle described in the first aspect of the present application.
[0009] The above technical solution provided by the embodiment of the present application has the following advantages over the prior art: the method provided by the embodiment of the present application firstly reduces the number of point clouds by downsampling the source point cloud set while retaining the better point clouds in the source point cloud set, thereby improving the speed of subsequent registration. In addition, in the embodiment of the present application, feature point detection is performed on the point clouds in the source point cloud set and the point clouds in the target point cloud, and then registration is performed to further improve the registration speed. Finally, in the embodiment of the present application, point pairs are found based on the Euclidean distance of the centroid, and different weights are assigned to point pairs with different Euclidean distances, thereby improving the registration accuracy while improving the accuracy of the error function value. Therefore, through the method of the embodiment of the present application, the registration speed and registration accuracy can be effectively improved, and the problem of low registration efficiency in the prior art for the registration task of point clouds of complex mechanical parts is solved. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0011] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0012] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.
[0013] Figure 1 A flow chart of a point cloud registration method for a vehicle steering knuckle provided in an embodiment of the present application;
[0014] Figure 2 A schematic diagram of a point cloud registration device for an automobile steering knuckle provided in an embodiment of the present application;
[0015] Figure 3 A schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0016] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0017] The disclosure below provides many different embodiments or examples to implement different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.
[0018] Figure 1 A flowchart of a point cloud registration method for a vehicle steering knuckle provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the steps of the method include:
[0019] Step 101, obtaining a source point cloud set and a target point cloud set corresponding to the steering knuckle, wherein the point clouds in the source point cloud set are point clouds obtained at different angles, and the point clouds in the target point cloud set are preset point clouds;
[0020] In a specific example, the collection of the source point cloud set and the target point cloud set can be: using structured light technology to obtain point clouds at different angles of the steering knuckle, let the source point cloud set be P = {p1, p2, ..., p m}, the target point cloud set is Q = {q1,q2,…,q n}.
[0021] Step 102, dividing the point cloud in the source point cloud set into voxel grids, and downsampling the point cloud in the source point cloud set based on the distance between the point in the voxel grid and the centroid in the voxel grid;
[0022] In the specific example, the steering node cloud is divided into a voxel grid, and for all points {p1,p2,…,p j}, calculate the mean of its X, Y, and Z coordinates, as shown in Formula 1, to form the three-dimensional coordinates of the center of mass of the current voxel The point with the smallest Euclidean distance from the centroid in the voxel is retained, and the remaining points are removed to achieve point cloud downsampling.
[0023]
[0024] The prior art uses voxel filtering and downsampling, that is, the voxel center point is used to replace all the points in the voxel. The center point is not necessarily the point cloud of the steering knuckle itself, which will reduce the accuracy of the point cloud model. In this application, the steering node closest to the voxel centroid is used to replace all the points in the voxel, which realizes point cloud downsampling and retains the point cloud of the steering knuckle itself without loss of accuracy.
[0025] Step 103, performing feature point detection on the downsampled source point cloud set and the target point cloud, and registering the downsampled source point cloud set and the target point cloud based on the feature points;
[0026] Step 104, determine multiple point cloud pairs from the downsampled source point cloud set and the target point cloud, set different weights for point cloud pairs with different Euclidean distances, and obtain the aligned point cloud set based on a preset error function, wherein the point cloud pair refers to a point cloud pair whose Euclidean distance between the first point cloud of the downsampled source point cloud set and the second point cloud in the target point cloud is less than a preset threshold.
[0027] Through the above steps 101 to 104, the source point cloud set is first downsampled to reduce the number of point clouds while retaining the better point clouds in the source point cloud set, thereby improving the speed of subsequent registration. In addition, in the embodiment of the present application, feature point detection is performed on the point clouds in the source point cloud set and the point clouds in the target point cloud, and then registration is performed to further improve the registration speed. Finally, in the embodiment of the present application, point pairs are found based on the centroid Euclidean distance, and different weights are assigned to point pairs with different Euclidean distances, thereby improving the registration accuracy while improving the accuracy of the error function value. Therefore, through the method of the embodiment of the present application, the registration speed and registration accuracy can be effectively improved, and the problem of low registration efficiency in the prior art for the registration task of complex mechanical parts point clouds is solved.
[0028] In an optional implementation of the embodiment of the present application, the method of performing feature point detection on the downsampled source point cloud set and the target point cloud involved in the above step 103, and registering the downsampled source point cloud set and the target point cloud based on the feature points, may further include:
[0029] Step 11, performing feature point detection on the downsampled source point cloud set and the target point cloud based on the K-4PCS algorithm to obtain a feature point set;
[0030] Step 12: align the downsampled source point cloud set and the target point cloud based on the feature point set.
[0031] In this regard, in the specific implementation of the embodiment of the present application, the K-4PCS (Keypoint-based FourPoints Congruent Sets) algorithm is used to first detect feature points of the point cloud, and then roughly align the source point cloud and the target point cloud based on the feature point set to provide a better initial value for precise alignment. The K-4PCS algorithm is selected as the point cloud coarse alignment method instead of the other commonly used SCA-IA (Sample Consensus Intial Alignment) algorithm and PCA (Principal Component Analysis) algorithm because K-4PCS calculates the transformation matrix based on the feature points of the point cloud, while SCA-IA and PCA calculate the transformation matrix based on the entire point cloud. Therefore, K-4PCS is the fastest and can meet the real-time requirements of the steering node cloud alignment. The above steps 11 and 12 are equivalent to performing a coarse alignment of the point cloud.
[0032] In an optional implementation of the embodiment of the present application, the method of determining a plurality of point cloud pairs from the downsampled source point cloud set and the target point cloud involved in step 104, setting different weights for point cloud pairs with different Euclidean distances, and obtaining the registered point cloud set based on a preset error function may further include:
[0033] Step 21, respectively obtaining the three-dimensional coordinate values of each point cloud in the downsampled source point cloud set and the target point cloud;
[0034] Step 22, based on the three-dimensional coordinate value, for each point cloud in the downsampled source point cloud set, a point cloud whose Euclidean distance is less than a preset threshold is determined from the target point cloud set as a corresponding point cloud, and the point cloud in the downsampled source point cloud set and the corresponding point cloud in the target point cloud set are determined as a point cloud pair;
[0035] Step 23, determining the weight of the point cloud pair based on the Euclidean distance of the point cloud pair, the average value of the Euclidean distance, and the variance of the Euclidean distance;
[0036] Step 24, performing a null position transformation on the downsampled source point cloud set based on the weight and the preset error function to obtain a registered point cloud set.
[0037] For the weight involved in the above step 23, the weight ω can be determined by the following formula:
[0038]
[0039] Among them, α is the average value of the Euclidean distance, s is the number of point cloud pairs, and σ 2 is the variance of the Euclidean distance, l i is the Euclidean distance between the i-th point cloud pair, ω i is the weight of the i-th point cloud pair.
[0040] The method of performing empty position transformation on the downsampled source point cloud set based on the weight and the preset error function to obtain the registered point cloud set involved in step 24 may further include:
[0041] Step 31, calculating the rotation matrix and translation matrix when the value of the preset error function is the smallest based on the least square method;
[0042] Step 32, performing a position transformation on the downsampled source point cloud set based on the rotation matrix and the translation matrix to obtain a registered point cloud set.
[0043] Among them, the formula of the preset error function is as shown in Formula 5:
[0044]
[0045] Among them, f is the error function, R is the rotation matrix, T is the translation matrix, s is the number of point cloud pairs, p i is the point cloud and q in the source point cloud set after downsampling i is a point cloud in the target point cloud set, and p i and q i is a point cloud pair.
[0046] The above steps 21 to 24 are for precise registration of point clouds relative to steps 11 and 12, and in a specific example, the following process may be included:
[0047] Step 41, for all points in the source point cloud and the target point cloud, the mean of their X, Y, and Z coordinates is calculated respectively. The calculation principle is the same as the calculation principle of the center of mass in the downsampling process, and the three-dimensional center of mass P0 (x p ,y p , z p ) and the 3D centroid Q0(x q ,y q , z q ), calculate the Euclidean distance d between the centroids, where d is as shown in Formula 6:
[0048]
[0049] Step 42: for each point in the source point cloud P, find the nearest point with a Euclidean distance less than d in the target point cloud Q as its corresponding point, and save all point pairs to form an overlapping area P1 = {p1, p2, ..., p s} and Q1={q1,q2,…,q s}.
[0050] Step 43, the number of point pairs in the overlapping area is s, and the Euclidean distance l between all point pairs, as well as their mean α and variance σ are calculated. 2 , the calculation method is shown in Formula 2 and Formula 3. For each point pair, different weights are assigned to it. The larger the Euclidean distance, the smaller the weight. The calculation method of weight ω is shown in Formula 4.
[0051] Step 44, using the point pairs in the overlapping area to construct an error function, and calculating the rotation matrix and translation matrix with the minimum error function value based on the least squares method, the error function expression is shown in Formula 5. The larger the Euclidean distance between the point pairs, the greater the registration error introduced. After assigning different weights by the method in this paper, the point pairs with larger Euclidean distances have smaller weights and lower contribution to the error function, which can effectively reduce the registration error. Finally, the weights of all point pairs are normalized to improve the accuracy of the error function value.
[0052] Step 45, using the rotation matrix R and translation matrix T calculated above to perform spatial position transformation on the source point cloud P, to obtain a transformed source point cloud.
[0053] Step 46, replacing the source point cloud before transformation with the transformed source point cloud, and repeating the above-mentioned precise registration steps until the error function value is less than the set threshold or the number of iterations reaches the maximum number of iterations.
[0054] It can be seen that the embodiment of the present application provides a fast registration method for the steering node cloud of an automobile, which uses structured light technology to collect steering node cloud data, and based on the innovative point cloud downsampling and fine registration method, realizes fast and accurate registration from coarse to fine, so as to reconstruct a high-precision three-dimensional model of the steering knuckle, provide accurate spatial position information for the end of the robotic arm, and ensure its accurate grasping. That is, in the point cloud downsampling stage, the Euclidean distance between each point and the centroid is compared, the number of point clouds is reduced while the best points are retained, which not only improves the speed of the registration process, but also suppresses the reduction of accuracy. In the point cloud coarse registration stage, registration is performed on the feature point set to improve the registration speed, and at the same time, a good initial value is provided for the fine registration to improve the registration accuracy. In the point cloud fine registration stage, point pairs are found based on the centroid Euclidean distance, and different weights are assigned to point pairs with different Euclidean distances, which improves the accuracy of the error function value and thus improves the registration accuracy. In addition, point pairs with a distance greater than the Euclidean distance between centroids are excluded, which reduces the amount of calculation and effectively improves the registration speed.
[0055] Corresponding to the above Figure 1 The present application also provides a point cloud registration device for a car steering knuckle. Figure 2 As shown, the device comprises:
[0056] An acquisition module 202 is used to acquire a source point cloud set and a target point cloud set corresponding to the steering knuckle, wherein the point clouds in the source point cloud set are point clouds acquired at different angles, and the point clouds in the target point cloud set are preset point clouds;
[0057] A first processing module 204 is used to divide the point cloud in the source point cloud set into voxel grids, and downsample the point cloud in the source point cloud set based on the distance between the point in the voxel grid and the centroid in the voxel grid;
[0058] The second processing module 206 is used to detect feature points on the downsampled source point cloud set and the target point cloud, and to register the downsampled source point cloud set and the target point cloud based on the feature points;
[0059] The third processing module 208 is used to determine multiple point cloud pairs from the downsampled source point cloud set and the target point cloud, set different weights for point cloud pairs with different Euclidean distances, and obtain the aligned point cloud set based on a preset error function, wherein the point cloud pair refers to the Euclidean distance between the first point cloud of the downsampled source point cloud set and the second point cloud in the target point cloud is less than a preset threshold.
[0060] Through the device of the embodiment of the present application, the source point cloud set is first downsampled to reduce the number of point clouds while retaining the better point clouds in the source point cloud set, thereby improving the speed of subsequent registration. In addition, in the embodiment of the present application, feature point detection is performed on the point clouds in the source point cloud set and the point clouds in the target point cloud, and then registration is performed to further improve the registration speed. Finally, in the embodiment of the present application, point pairs are found based on the centroid Euclidean distance, and different weights are assigned to point pairs with different Euclidean distances, thereby improving the registration accuracy while improving the accuracy of the error function value. Therefore, through the method of the embodiment of the present application, the registration speed and registration accuracy can be effectively improved, solving the problem of low registration efficiency in the prior art for the registration task of complex mechanical parts point clouds.
[0061] In an optional implementation of the embodiment of the present application, the second processing module 206 in the embodiment of the present application may further include: a detection unit, used to perform feature point detection on the downsampled source point cloud set and the target point cloud based on the K-4PCS algorithm to obtain a feature point set; a first alignment unit, used to align the downsampled source point cloud set and the target point cloud based on the feature point set.
[0062] In an optional implementation manner of the embodiment of the present application, the third processing module in the embodiment of the present application may further include: an acquisition unit, used to respectively acquire the three-dimensional coordinate values of each point cloud in the downsampled source point cloud set and the target point cloud; a first processing unit, used to determine, based on the three-dimensional coordinate values, for each point cloud in the downsampled source point cloud set, a point cloud whose Euclidean distance is less than a preset threshold from the target point cloud set as the corresponding point cloud, and determine the point cloud in the downsampled source point cloud set and the corresponding point cloud in the target point cloud set as a point cloud pair; a second processing unit, used to determine the weight of the point cloud pair based on the Euclidean distance of the point cloud pair, the average value of the Euclidean distance and the variance of the Euclidean distance; a second registration unit, used to perform an empty position transformation on the downsampled source point cloud set based on the weight and a preset error function to obtain a registered point cloud set.
[0063] In an optional implementation of the embodiment of the present application, the weight ω is determined by the following formula:
[0064]
[0065] Among them, α is the average value of the Euclidean distance, s is the number of point cloud pairs, and σ 2 is the variance of the Euclidean distance, l i is the Euclidean distance between the i-th point cloud pair, ω i is the weight of the i-th point cloud pair.
[0066] In an optional implementation manner of an embodiment of the present application, the above-mentioned second registration unit may further include: a calculation subunit, used to calculate the rotation matrix and translation matrix when the value of a preset error function is minimized based on the least squares method; a registration subunit, used to perform empty position transformation on the downsampled source point cloud set based on the rotation matrix and translation matrix to obtain a registered point cloud set.
[0067] In an optional implementation manner of the embodiment of the present application, the formula of the preset error function is:
[0068]
[0069] Among them, f is the error function, R is the rotation matrix, T is the translation matrix, s is the number of point cloud pairs, p i is the point cloud and q in the source point cloud set after downsampling i is a point cloud in the target point cloud set, and p i and q i is a point cloud pair.
[0070] like Figure 3 As shown, an embodiment of the present application provides an electronic device, including a processor 311, a communication interface 312, a memory 313 and a communication bus 314, wherein the processor 311, the communication interface 312, and the memory 313 communicate with each other through the communication bus 314.
[0071] Memory 313, used for storing computer programs;
[0072] In one embodiment of the present application, the processor 311 is used to execute the program stored in the memory 313 to implement the point cloud registration method of the automobile steering knuckle provided by any of the aforementioned method embodiments, and the role it plays is similar and will not be repeated here.
[0073] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the point cloud registration method for a vehicle steering knuckle provided in any of the aforementioned method embodiments are implemented.
[0074] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0075] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0076] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.
[0077] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.
Claims
1. A point cloud registration method for a car steering knuckle, characterized in that: include: Acquire a source point cloud set and a target point cloud set corresponding to the steering knuckle, wherein the point clouds in the source point cloud set are point clouds acquired at different angles, and the point clouds in the target point cloud set are preset point clouds; Dividing the point cloud in the source point cloud set into a voxel grid, and downsampling the point cloud in the source point cloud set based on the distance between the point in the voxel grid and the centroid in the voxel grid; Performing feature point detection on the downsampled source point cloud set and the target point cloud, and registering the downsampled source point cloud set and the target point cloud based on the feature points; A plurality of point cloud pairs are determined from the downsampled source point cloud set and the target point cloud, different weights are set for point cloud pairs with different Euclidean distances, and a registered point cloud set is obtained based on a preset error function, wherein the point cloud pair refers to a point cloud pair whose Euclidean distance between a first point cloud of the downsampled source point cloud set and a second point cloud in the target point cloud is less than a preset threshold.
2. The method according to claim 1, characterized in that: Performing feature point detection on the downsampled source point cloud set and the target point cloud, and registering the downsampled source point cloud set and the target point cloud based on the feature points, including: Based on the K-4PCS algorithm, feature point detection is performed on the downsampled source point cloud set and the target point cloud to obtain a feature point set; The downsampled source point cloud set and the target point cloud are registered based on the feature point set.
3. The method according to claim 1, characterized in that A plurality of point cloud pairs are determined from the downsampled source point cloud set and the target point cloud, different weights are set for point cloud pairs with different Euclidean distances, and the aligned point cloud set is obtained based on a preset error function, including: Respectively obtaining the three-dimensional coordinate values of each point cloud in the downsampled source point cloud set and the target point cloud; Based on the three-dimensional coordinate values, for each point cloud in the downsampled source point cloud set, a point cloud having a Euclidean distance less than a preset threshold is determined from the target point cloud set as a corresponding point cloud, and the point cloud in the downsampled source point cloud set and the corresponding point cloud in the target point cloud set are determined as the point cloud pair; Determining a weight of the point cloud pair based on the Euclidean distance of the point cloud pair, an average value of the Euclidean distance, and a variance of the Euclidean distance; Based on the weight and the preset error function, the downsampled source point cloud set is transformed in empty position to obtain a registered point cloud set.
4. The method according to claim 3, characterized in that The weight ω is determined by the following formula: Among them, α is the average value of the Euclidean distance, s is the number of point cloud pairs, and σ 2 is the variance of the Euclidean distance, l i is the Euclidean distance between the i-th point cloud pair, ω i is the weight of the i-th point cloud pair.
5. The method according to claim 1, characterized in that: Performing a null position transformation on the downsampled source point cloud set based on the weight and the preset error function to obtain a registered point cloud set includes: Calculate the rotation matrix and translation matrix when the value of the preset error function is minimized based on the least squares method; The downsampled source point cloud set is transformed in empty position based on the rotation matrix and the translation matrix to obtain a registered point cloud set.
6. The method according to claim 5, characterized in that The formula of the preset error function is: Among them, f is the error function, R is the rotation matrix, T is the translation matrix, s is the number of point cloud pairs, p i is the point cloud and q in the source point cloud set after downsampling i is a point cloud in the target point cloud set, and p i and q i is a point cloud pair.
7. A point cloud registration device for a car steering knuckle, characterized in that: include: An acquisition module, used to acquire a source point cloud set and a target point cloud set corresponding to the steering knuckle, wherein the point clouds in the source point cloud set are point clouds acquired at different angles, and the point clouds in the target point cloud set are preset point clouds; A first processing module, configured to divide the point cloud in the source point cloud set into voxel grids, and downsample the point cloud in the source point cloud set based on the distance between the point in the voxel grid and the centroid in the voxel grid; A second processing module is used to detect feature points on the downsampled source point cloud set and the target point cloud, and to register the downsampled source point cloud set and the target point cloud based on the feature points; The third processing module is used to determine multiple point cloud pairs from the downsampled source point cloud set and the target point cloud, set different weights for point cloud pairs with different Euclidean distances, and obtain the aligned point cloud set based on a preset error function, wherein the point cloud pair refers to the Euclidean distance between the first point cloud of the downsampled source point cloud set and the second point cloud in the target point cloud is less than a preset threshold.
8. The device according to claim 7, characterized in that The third processing module includes: An acquisition unit, used to respectively acquire the three-dimensional coordinate values of each point cloud in the downsampled source point cloud set and the target point cloud; A first processing unit is configured to determine, for each point cloud in the downsampled source point cloud set based on the three-dimensional coordinate value, a point cloud having a Euclidean distance less than a preset threshold from the target point cloud set as a corresponding point cloud, and determine the point cloud in the downsampled source point cloud set and the corresponding point cloud in the target point cloud set as the point cloud pair; A second processing unit, configured to determine a weight of the point cloud pair based on the Euclidean distance of the point cloud pair, an average value of the Euclidean distance, and a variance of the Euclidean distance; The second registration unit is used to perform a null position transformation on the downsampled source point cloud set based on the weight and the preset error function to obtain a registered point cloud set.
9. An electronic device, comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor coupled to the at least one bus; At least one memory connected to the at least one bus, wherein the processor is configured to execute the point cloud registration method for the automobile steering knuckle as described in any one of claims 1 to 6 of the present application.
10. A computer storage medium storing computer executable instructions, wherein the computer executable instructions are used to execute the point cloud registration method for an automobile steering knuckle as described in any one of claims 1 to 6 of the present application.