Automatic combined point cloud registration system and method
Through automated combination of point cloud registration system, using parameter configuration, preprocessing and iterative optimization algorithms, the problem of point cloud registration in the existing technology requires manual intervention and time-consuming, and the rapid and accurate data of complex scene point clouds is achieved. Combination improves work efficiency and data quality.
Patent Information
- Application Number
- CN202410832974.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-26
- Publication Date
- 2024-11-01
Smart Images

Figure CN118887263A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of automatic registration, and more specifically, to an automatic combined point cloud registration system and method. Background Art
[0002] With the rapid development of science and technology, 3D reconstruction technology has demonstrated its unique value in many fields, such as autonomous driving, robot navigation, and medical image analysis. In these applications, point cloud data, as an important form of representation of three-dimensional scenes, is crucial to building high-quality three-dimensional models with its accuracy and integrity. However, due to the limitations of the sensor's viewing angle and the complexity of the environment, a single scan can often only obtain partial point cloud data on the surface of the target object. Therefore, how to effectively combine these scattered point cloud data to form a complete and consistent three-dimensional model has become an important challenge in the field of computer vision and three-dimensional reconstruction.
[0003] Point cloud registration technology is the key to solving this challenge. Simply put, point cloud registration is to align multiple point cloud data sets into a unified coordinate system to obtain an overall consistent point cloud model. This technology not only requires the ability to accurately calculate the spatial transformation relationship between different point clouds, but also needs to consider the influence of factors such as noise, occlusion, and incomplete data. Through point cloud registration, we can fuse the point cloud data obtained from multiple scans to form a more complete and accurate 3D model, providing strong data support for subsequent 3D reconstruction, object recognition and tracking and other applications.
[0004] However, it still has some shortcomings in actual use. For example, traditional point cloud registration methods usually rely on human intervention, requiring manual calibration of the collected point cloud data or manual selection of registration feature points, and the registration process is time-consuming, labor-intensive and unstable. Summary of the invention
[0005] In order to overcome the above defects of the prior art, the embodiments of the present invention provide x to solve the problems raised in the above background technology.
[0006] To achieve the above object, the present invention provides the following technical solutions:
[0007] Fill in parameter configuration module: used to fill in parameter configuration files according to actual needs. By setting a large number of parameters, you can customize a variety of combined point cloud registration methods for various environments;
[0008] Input point cloud module: used to support the input of multiple PCD and frame files for point cloud registration into a complete point cloud;
[0009] Point cloud preprocessing module: used to preprocess the input point cloud, reduce the number of input point clouds and improve calculation efficiency;
[0010] Feature extraction module: used to downsample the source point cloud and the target point cloud and extract features to further reduce the amount of calculation and improve the registration accuracy;
[0011] Rough registration module: It is used to perform rough registration first and output a rough transformation matrix to improve the initial position conditions, avoid falling into the local optimal situation, and improve the accuracy of fine registration;
[0012] Precision matching module: used for precise matching and output of the final transformation matrix;
[0013] Target point cloud merging module: used to merge the registered source point cloud and target point cloud and then perform downsampling to output the point cloud.
[0014] Preferably, in the parameter configuration module, the parameter configuration file is filled in according to actual needs, and a variety of combined point cloud registration methods can be customized by setting a large number of parameters, so as to be suitable for applications in various environments. The parameter configuration file may include the format of point cloud data, the selection of registration algorithm, the accuracy requirements of registration, and other related parameter settings. Through flexible parameter configuration, users can select the appropriate registration method according to specific needs and environmental characteristics, so as to obtain more accurate and stable registration results.
[0015] Preferably, in the input point cloud module, the function of supporting the input of multiple PCD or frame files for point cloud registration into a complete point cloud can be realized through a parameter configuration file. The user can specify the format and path of the point cloud data file to be registered, as well as the registration method and parameter settings in the parameter configuration file; through the configuration file, the user can set the number, order and processing method of the input files, and then register multiple point cloud files to obtain complete point cloud data;
[0016] In the parameter configuration file, users can set the format and path of the input file, specify the file name and path of the point cloud data to be registered. At the same time, users can also specify the selection of the registration method and related parameters, such as the number of iterations, convergence accuracy, feature descriptors, and the point cloud fusion method after registration.
[0017] Preferably, in the point cloud preprocessing module, denoising removes noise points and outliers in the point cloud data; these noise points are caused by scanning equipment, environment or other factors;
[0018] Commonly used denoising algorithms include statistical filtering, median filtering, Gaussian filtering, etc. Among them, statistical filtering can remove abnormal points based on the statistical characteristics of point cloud data; median filtering removes scattered noise by replacing each point with the median of its neighborhood; Gaussian filtering uses Gaussian distribution as a neighborhood filter with weights, which can smooth the image while retaining the edges;
[0019] Voxel downsampling is a method of downsampling point cloud data. The basic principle is to divide the point cloud data into small voxels, where small voxels are small cube units, and only retain one point inside each voxel, while all remaining points are discarded. This method can reduce the size of point cloud data, thereby reducing the amount of calculation and improving the speed and efficiency of point cloud data processing.
[0020] Preferably, in the feature extraction module, curvature-based downsampling is used for sampling, wherein the curvature-based downsampling method is specifically:
[0021] The downsampling process is guided by the local curvature of the point cloud, retaining more points in areas with large curvature changes and reducing the number of points in flat areas;
[0022] For each point in the point cloud, calculate its curvature information. Curvature represents the degree of curvature of the surface and is an important measure of the local geometric characteristics of the point cloud. The curvature is estimated by calculating the K neighborhood of each point and then analyzing the relationship between these neighboring points and the point.
[0023] According to application requirements, a curvature threshold is set; points above this threshold are considered to be areas with obvious features, while points below this threshold are considered to be areas with unclear features;
[0024] The choice of curvature threshold directly affects the downsampling result and needs to be adjusted according to the actual situation;
[0025] According to the calculated curvature information and the set threshold, the point cloud is divided into regions with obvious features and regions with unclear features. Different sampling strategies are used for these two regions. For regions with obvious features, more points can be retained to preserve details; while for regions with unclear features, the number of points can be reduced to reduce data density.
[0026] The specific sampling strategy can be customized according to actual needs, and sampling can be carried out according to a fixed ratio or quantity; after completing the initial underground sampling, some optimization and post-processing steps are performed to smooth the sampled point cloud to reduce the impact of noise and outliers.
[0027] Preferably, in the rough registration module, during the rough registration process, the output rough transformation matrix usually includes a rotation matrix R and a translation vector t. The rotation matrix R is a 3x3 matrix representing a rotation in a three-dimensional space; the translation vector t is a 3x1 matrix representing a translation in a three-dimensional space.
[0028] Preferably, in the fine matching module, the fine matching module first receives the initial transformation matrix output by the rough matching as a starting point; the initial transformation matrix provides a rough alignment basis for the fine matching;
[0029] The precision matching module uses an iterative optimization algorithm and an iterative closest point algorithm to perform alignment optimization. In each round of iteration, it performs the following steps:
[0030] Corresponding point matching: Based on the current transformation matrix, each point in the source point cloud is mapped to the target point cloud and the nearest corresponding point is found; these corresponding point pairs are used to calculate the new transformation matrix;
[0031] Transformation matrix calculation: Use the corresponding point pairs to calculate a new transformation matrix through the least square method or other optimization methods. This new transformation matrix can more accurately describe the alignment relationship between the source point cloud and the target point cloud;
[0032] Update the transformation matrix: use the new transformation matrix as the current transformation matrix and prepare for the next round of iteration;
[0033] Convergence judgment: After each round of iteration, the fine matching module will check whether the preset convergence conditions are met, such as reaching the maximum number of iterations, alignment error less than the threshold, etc. If the convergence conditions are met, the iteration is stopped and the final transformation matrix is output; otherwise, the next round of iteration will continue.
[0034] Result output: When the iteration converges, the fine matching module outputs the final transformation matrix. This matrix includes a rotation matrix and a translation vector, which can accurately transform the source point cloud to a position aligned with the target point cloud.
[0035] Preferably, in the target point cloud merging module, before merging, the registration result needs to be verified to ensure that the source point cloud has been accurately aligned with the target point cloud; by calculating the alignment error between the two point clouds, observing the matching degree of the overlapping area or using visualization tools for verification; once the registration result meets the requirements, the source point cloud and the target point cloud can be merged into a single point cloud data; the merging operation transforms each point in the source point cloud into the coordinate system of the target point cloud according to the transformation matrix calculated during the registration process, and then merges the transformed points with the original target point cloud; during the merging process, there may be overlapping areas between the source point cloud and the target point cloud. The points in these overlapping areas may be redundant, and these redundant points can be eliminated by using a distance-based clustering algorithm to identify and delete duplicate points in the overlapping areas.
[0036] Technical effects and advantages of the present invention:
[0037] 1. Through automated processing, the system can quickly register and combine point cloud data of large-scale and complex scenes, greatly improving work efficiency;
[0038] 2. The system uses advanced algorithms and computer technology to achieve accurate registration and combination of point cloud data, ensuring the data quality of subsequent tasks. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 Schematic diagram of system module connection of the present invention
[0040] Figure 2 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION
[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0042] See also Figure 1 As shown, the present invention relates to an automated combined point cloud registration system, which includes a parameter configuration module, a point cloud input module, a point cloud preprocessing module, a feature extraction module, a rough matching module, a fine matching module, and a target point cloud merging module.
[0043] The parameter filling configuration module is connected to the input point cloud module, the input point cloud module is connected to the point cloud preprocessing module, the point cloud preprocessing module is connected to the feature extraction module, the feature extraction module is connected to the rough matching module, the rough matching module is connected to the fine matching module, and the fine matching module is connected to the target point cloud merging module.
[0044] The parameter configuration module is used to fill in the parameter configuration file according to actual needs. By setting a large number of parameters, you can customize a variety of combined point cloud registration methods for various different environments;
[0045] In the parameter configuration module, fill in the parameter configuration file according to actual needs. You can customize multiple combined point cloud registration methods by setting a large number of parameters, so that they are suitable for applications in various environments. The parameter configuration file can include the format of point cloud data, the selection of registration algorithm, the accuracy requirements of registration, and other related parameter settings. Through flexible parameter configuration, users can choose the appropriate registration method according to specific needs and environmental characteristics, so as to obtain more accurate and stable registration results;
[0046] In the parameter configuration file, users can set the parameters of the registration method, such as the number of iterations, convergence accuracy, selection of local feature descriptors, etc., and adjust these parameters according to actual conditions. In addition, users can also set the processing method of point cloud data, such as filtering, denoising, feature extraction, etc., as well as the evaluation indicators after registration, such as overlap, error value, etc. The setting of these parameters can help users achieve more accurate point cloud registration in different environments; through the parameter configuration file, users can flexibly select and adjust the registration method and parameters according to specific application scenarios and needs, so as to obtain more accurate and efficient point cloud registration results.
[0047] The input point cloud module is used to support input of multiple PCD and frame files for point cloud registration into a complete point cloud;
[0048] In the input point cloud module, the function of supporting the input of multiple PCD or frame files for point cloud registration into a complete point cloud can be realized through the parameter configuration file. The user can specify the format and path of the point cloud data file to be registered, as well as the registration method and parameter settings in the parameter configuration file; through the configuration file, the user can set the number, order and processing method of the input files, and then register multiple point cloud files to obtain complete point cloud data;
[0049] In the parameter configuration file, users can set the format and path of the input file, specify the file name and path of the point cloud data to be registered. At the same time, users can also specify the selection of the registration method and related parameters, such as the number of iterations, convergence accuracy, feature descriptors, and the point cloud fusion method after registration.
[0050] Through the flexible setting of parameter configuration files, users can flexibly adjust the registration method and parameters according to actual needs and environmental characteristics, thereby realizing the registration and fusion of multiple point cloud files. This functional expansion provides users with a more flexible and convenient point cloud processing method, which can be applied to a variety of different scenarios, such as map construction, 3D reconstruction and other fields, providing users with more choices and optimization possibilities. At the same time, the use of parameter configuration files also improves the customizability and universality of the point cloud registration method, making it more flexible and practical in practical applications.
[0051] The point cloud preprocessing module preprocesses the input point cloud to reduce the number of input point clouds and improve calculation efficiency;
[0052] In the point cloud preprocessing module, denoising removes noise points and outliers in the point cloud data; these noise points are caused by scanning equipment, environment or other factors;
[0053] Commonly used denoising algorithms include statistical filtering, median filtering, Gaussian filtering, etc. Among them, statistical filtering can remove abnormal points based on the statistical characteristics of point cloud data; median filtering removes scattered noise by replacing each point with the median of its neighborhood; Gaussian filtering uses Gaussian distribution as a neighborhood filter with weights, which can smooth the image while retaining the edges;
[0054] Voxel downsampling is a method of downsampling point cloud data. Its basic principle is to divide the point cloud data into small voxels, where small voxels are small cube units, and only one point inside each voxel is retained, while all the remaining points are discarded. This method can reduce the size of point cloud data, thereby reducing the amount of calculation and improving the speed and efficiency of point cloud data processing.
[0055] Coordinate axis alignment Due to multi-view scanning or other reasons, point cloud data may be defined in their respective local coordinate systems; for subsequent processing and analysis, these point cloud data need to be converted to a unified global coordinate system;
[0056] The method of point cloud alignment includes using a measuring device to automatically record the transformation between viewing angles, aligning point clouds by calculating the topological relationship between landmark points, and converting the point cloud in the local coordinate system to the global coordinate system through the coordinate transformation relationship of the overlapping part.
[0057] The feature extraction module is used to downsample the source point cloud and the target point cloud and extract features to further reduce the amount of calculation and improve the registration accuracy;
[0058] In the feature extraction module, curvature-based downsampling is used for sampling, wherein the curvature-based downsampling method is specifically as follows:
[0059] The downsampling process is guided by the local curvature of the point cloud, retaining more points in areas with large curvature changes and reducing the number of points in flat areas;
[0060] For each point in the point cloud, calculate its curvature information. Curvature represents the degree of curvature of the surface and is an important measure of the local geometric characteristics of the point cloud. The curvature is estimated by calculating the K neighborhood of each point and then analyzing the relationship between these neighboring points and the point.
[0061] The calculation method for calculating curvature using least squares fitting is as follows:
[0062] Z=ax 2 +bxy+cy 2 , where a, b, and c are constants that determine the shape and size of the surface; x and y are the parameters of the surface, and z is the corresponding vertical distance; ax 2 +cy 2describes the quadratic term in the x and y directions, while bxy is the cross term, which describes the interaction between the x and y directions.
[0063] According to application requirements, a curvature threshold is set; points above this threshold are considered to be areas with obvious features, while points below this threshold are considered to be areas with unclear features;
[0064] The choice of curvature threshold directly affects the downsampling result and needs to be adjusted according to the actual situation;
[0065] According to the calculated curvature information and the set threshold, the point cloud is divided into regions with obvious features and regions with unclear features. Different sampling strategies are used for these two regions. For regions with obvious features, more points can be retained to preserve details; while for regions with unclear features, the number of points can be reduced to reduce data density.
[0066] The specific sampling strategy can be customized according to actual needs, and sampling can be carried out according to a fixed ratio or quantity; after completing the initial underground sampling, some optimization and post-processing steps are performed to smooth the sampled point cloud to reduce the impact of noise and outliers.
[0067] The rough registration module is used to first perform rough registration to output a rough transformation matrix, improve the initial position conditions, avoid falling into a local optimal situation, and improve the accuracy of fine registration;
[0068] In the rough registration module, during the rough registration process, the output rough transformation matrix usually includes a rotation matrix R and a translation vector t. The rotation matrix R is a 3x3 matrix, which represents the rotation in three-dimensional space; the translation vector t is a 3x1 matrix, which represents the translation in three-dimensional space. The calculation method of the rotation matrix is as follows:
[0069] Among them, r ij are the elements of the matrix, and R is an orthogonal matrix, that is, R T *R=I, I represents the identity matrix, and det(R)=1;
[0070] The calculation method of the translation amount t is as follows:
[0071] Among them, t x Expressed as the translation on the x-axis, t y Expressed as the translation on the y-axis, t z Expressed as the amount of translation on the z-axis;
[0072] In the coarse registration process, the RANSAC algorithm is used to find corresponding matching point pairs, and the rotation matrix R and translation vector t are estimated based on these matching point pairs.
[0073] The precise alignment module is used to perform precise alignment and output a final conversion matrix;
[0074] In the fine matching module, the fine matching module first receives the initial transformation matrix output by the rough matching as a starting point; this initial transformation matrix provides a rough alignment basis for the fine matching;
[0075] The precision matching module uses an iterative optimization algorithm and an iterative closest point algorithm to perform alignment optimization. In each round of iteration, it performs the following steps:
[0076] Corresponding point matching: Based on the current transformation matrix, each point in the source point cloud is mapped to the target point cloud and the nearest corresponding point is found; these corresponding point pairs are used to calculate the new transformation matrix;
[0077] The distance between corresponding points is calculated as follows:
[0078] e i =‖R*g i +th i ‖, where e i It is represented as the distance between corresponding points, R is represented as the rotation matrix, t is represented as the translation vector, g i and h i Represented as corresponding points;
[0079] Transformation matrix calculation: Use the corresponding point pairs to calculate a new transformation matrix through the least square method or other optimization methods. This new transformation matrix can more accurately describe the alignment relationship between the source point cloud and the target point cloud;
[0080] Update the transformation matrix: use the new transformation matrix as the current transformation matrix and prepare for the next round of iteration;
[0081] Convergence judgment: After each round of iteration, the fine matching module will check whether the preset convergence conditions are met, such as reaching the maximum number of iterations, alignment error less than the threshold, etc. If the convergence conditions are met, the iteration is stopped and the final transformation matrix is output; otherwise, the next round of iteration will continue.
[0082] Result output: When the iteration converges, the fine matching module outputs the final transformation matrix. This matrix includes a rotation matrix and a translation vector, which can accurately transform the source point cloud to a position aligned with the target point cloud.
[0083] The target point cloud merging module merges the registered source point cloud and the target point cloud and then performs downsampling processing to output the point cloud;
[0084] In the target point cloud merging module, before merging, the registration results need to be verified to ensure that the source point cloud has been accurately aligned with the target point cloud; by calculating the alignment error between the two point clouds, observing the matching degree of the overlapping area or using visualization tools for verification; once the registration results meet the requirements, the source point cloud and the target point cloud can be merged into a single point cloud data; the merging operation transforms each point in the source point cloud into the coordinate system of the target point cloud according to the transformation matrix calculated during the registration process, and then merges the transformed points with the original target point cloud; during the merging process, there may be overlapping areas between the source point cloud and the target point cloud. The points in these overlapping areas may be redundant, and to eliminate these redundant points, a distance-based clustering algorithm can be used to identify and delete duplicate points in the overlapping areas.
[0085] See also Figure 2 As shown, in this embodiment, it should be specifically explained that the present invention provides an automatic combined point cloud registration method, comprising the following steps:
[0086] Step A1: Fill in the parameter configuration file according to actual needs. By setting a large number of parameters, you can customize a variety of combined point cloud registration methods for various environments;
[0087] Step A2: Support input of multiple PCD and frame files for point cloud registration into a complete point cloud;
[0088] Step A3: pre-process the input point cloud to reduce the number of input point clouds and improve calculation efficiency;
[0089] Step A4: downsample the source point cloud and the target point cloud and extract features to further reduce the amount of calculation and improve the registration accuracy;
[0090] Step A5: Perform a rough registration to output a rough transformation matrix, improve the initial position conditions, avoid falling into the local optimal situation, and improve the accuracy of the fine registration;
[0091] Step A6: Used to perform precise registration and output the final transformation matrix;
[0092] Step A7: Merge the registered source point cloud and target point cloud and then perform downsampling to output the point cloud.
[0093] Finally: 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 spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. Automated combined point cloud registration system, characterized by: include: Fill in parameter configuration module: used to fill in parameter configuration files according to actual needs. By setting a large number of parameters, you can customize a variety of combined point cloud registration methods for various environments; Input point cloud module: used to support the input of multiple PCD and frame files for point cloud registration into a complete point cloud; Point cloud preprocessing module: used to preprocess the input point cloud, reduce the number of input point clouds and improve calculation efficiency; Feature extraction module: used to downsample the source point cloud and the target point cloud and extract features to further reduce the amount of calculation and improve the registration accuracy; Rough registration module: used to perform rough registration first to output a rough transformation matrix, improve the initial position conditions, avoid falling into the local optimal situation, and improve the accuracy of fine registration; Precision matching module: used for precise matching and output of the final transformation matrix; Target point cloud merging module: used to merge the registered source point cloud and target point cloud and then perform downsampling to output the point cloud.
2. The automated combined point cloud registration system according to claim 1, characterized in that: In the parameter configuration module, the parameter configuration file is filled in according to actual needs. A variety of combined point cloud registration methods can be customized by setting a large number of parameters, so that they are suitable for applications in various environments; the parameter configuration file can include the format of point cloud data, the selection of registration algorithm, the accuracy requirements of registration, and other related parameter settings; through flexible parameter configuration, users can choose the appropriate registration method according to specific needs and environmental characteristics, so as to obtain more accurate and stable registration results.
3. The automated combined point cloud registration system according to claim 1, characterized in that: In the input point cloud module, the function of supporting the input of multiple PCD or frame files for point cloud registration into a complete point cloud can be realized through a parameter configuration file; the user can specify the format and path of the point cloud data file to be registered, as well as the registration method and parameter settings in the parameter configuration file; through the configuration file, the user can set the number, order and processing method of the input files, and then register multiple point cloud files to obtain complete point cloud data.
4. The automated combined point cloud registration system according to claim 1, characterized in that: In the point cloud preprocessing module, denoising removes noise points and outliers in the point cloud data; these noise points are caused by scanning equipment, environment or other factors; Voxel downsampling is a method of downsampling point cloud data; The basic principle is to divide the point cloud data into small voxels, where the small voxels are small cube units, and only one point inside each voxel is retained, while all the remaining points are discarded; this method can reduce the size of the point cloud data, thereby reducing the amount of calculation and improving the speed and efficiency of point cloud data processing; Coordinate axis alignment Due to multi-view scanning or other reasons, point cloud data may be defined in their respective local coordinate systems; for subsequent processing and analysis, these point cloud data need to be converted to a unified global coordinate system.
5. The automated combined point cloud registration system according to claim 1, characterized in that: In the feature extraction module, curvature-based downsampling is used for sampling, wherein the curvature-based downsampling method is specifically as follows: The downsampling process is guided by the local curvature of the point cloud, retaining more points in areas with large curvature changes and reducing the number of points in flat areas; For each point in the point cloud, calculate its curvature information; curvature represents the degree of curvature of the surface and is an important measure of the local geometric characteristics of the point cloud; the curvature is estimated by calculating the K neighborhood of each point and then analyzing the relationship between these neighboring points and the point; According to application requirements, a curvature threshold is set; points above this threshold are considered to be areas with obvious features, while points below this threshold are considered to be areas with unclear features; The choice of curvature threshold directly affects the downsampling result and needs to be adjusted according to the actual situation; According to the calculated curvature information and the set threshold, the point cloud is divided into areas with obvious features and areas with unclear features. Different sampling strategies are used for these two areas. For areas with obvious features, more points can be retained to preserve details; while for areas with unclear features, the number of points can be reduced to reduce data density.
6. The automated combined point cloud registration system according to claim 5, characterized in that: The calculation method for calculating curvature using least squares fitting is as follows: Z=ax 2 +bxy+cy 2 , where a, b, and c are constants that determine the shape and size of the surface; x and y are the parameters of the surface, and z is the corresponding vertical distance; ax 2 +cy 2 describes the quadratic term in the x and y directions, while bxy is the cross term, which describes the interaction between the x and y directions.
7. The automated combined point cloud registration system according to claim 1, characterized in that: In the rough registration module, during the rough registration process, the output rough transformation matrix usually includes a rotation matrix R and a translation vector t; the rotation matrix R is a 3x3 matrix, representing the rotation in the three-dimensional space; the translation vector t is a 3x1 matrix, representing the translation in the three-dimensional space.
8. The automated combined point cloud registration system according to claim 7, characterized in that: The calculation method of the rotation matrix is specifically as follows: in the fine matching module, the fine matching module first receives the initial transformation matrix output by the rough matching as a starting point; This initial transformation matrix provides a rough alignment basis for fine registration; The precision matching module uses an iterative optimization algorithm and an iterative closest point algorithm to perform alignment optimization. In each round of iteration, it performs the following steps: Corresponding point matching: Based on the current transformation matrix, each point in the source point cloud is mapped to the target point cloud and the nearest corresponding point is found; these corresponding point pairs are used to calculate the new transformation matrix; Transformation matrix calculation: Using the corresponding point pairs, a new transformation matrix is calculated by the least square method or other optimization methods; this new transformation matrix can more accurately describe the alignment relationship between the source point cloud and the target point cloud; Update the transformation matrix: use the new transformation matrix as the current transformation matrix and prepare for the next round of iteration; Convergence judgment: After each round of iteration, the fine matching module will check whether the preset convergence conditions are met, such as reaching the maximum number of iterations, alignment error less than the threshold, etc. If the convergence conditions are met, the iteration will be stopped and the final transformation matrix will be output; otherwise, the next round of iteration will be continued; Result output: When the iteration converges, the precision matching module outputs the final transformation matrix; this matrix includes a rotation matrix and a translation vector, which can accurately transform the source point cloud to a position aligned with the target point cloud.
9. The automated combined point cloud registration system according to claim 1, characterized in that: In the target point cloud merging module, the registration result needs to be verified before merging to ensure that the source point cloud has been accurately aligned with the target point cloud; By calculating the alignment error between the two point clouds, observing the degree of matching in the overlapping area or using visualization tools for verification; once the registration results meet the requirements, the source point cloud and the target point cloud can be merged into a single point cloud data; the merging operation transforms each point in the source point cloud into the coordinate system of the target point cloud according to the transformation matrix calculated during the registration process, and then merges the transformed points with the original target point cloud; during the merging process, there may be overlapping areas between the source point cloud and the target point cloud; the points in these overlapping areas may be redundant, and to eliminate these redundant points, a distance-based clustering algorithm can be used to identify and delete duplicate points in the overlapping areas.
10. An automated combined point cloud registration method, using the automated combined point cloud registration system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step A1: Fill in the parameter configuration file according to actual needs. By setting a large number of parameters, you can customize a variety of combined point cloud registration methods for various environments; Step A2: Support input of multiple PCD and frame files for point cloud registration into a complete point cloud; Step A3: pre-process the input point cloud to reduce the number of input point clouds and improve calculation efficiency; Step A4: downsample the source point cloud and the target point cloud and extract features to further reduce the amount of calculation and improve the registration accuracy; Step A5: Perform a rough registration to output a rough transformation matrix, improve the initial position conditions, avoid falling into the local optimal situation, and improve the accuracy of the fine registration; Step A6: Used to perform precise registration and output the final transformation matrix; Step A7: Merge the registered source point cloud and target point cloud and then perform downsampling to output the point cloud.
Citation Information
Cited By
Three-dimensional image reconstruction method and device
CN119417992A
Two-stage automatic registration method for brain magnetic measurement helmet and brain MRI data
CN120259392A
