A point cloud registration method and system based on a multi-feature point set

Abstract

The application relates to the field of point cloud registration, and provides a point cloud registration method and system based on a multi-feature point point set. The method comprises the following steps: preprocessing point cloud by statistical noise reduction based on Gaussian distribution and voxel downsampling; extracting multi-feature points by adopting a directional bounding box algorithm, an improved internal morphological descriptor algorithm and a geometric centroid point algorithm; constructing a multi-feature point point set; and obtaining an initial transformation matrix by adopting a principal component analysis-based coarse registration algorithm; performing fine registration by adopting a G-ICP algorithm based on L2LOSS robust loss function optimization; and outputting a registration result. The application improves the accuracy and efficiency of point cloud matching and reduces the registration time.

Description

Technical Field

[0001] This invention relates to the field of point cloud registration, and in particular to a point cloud registration method and system based on a multi-feature point set. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] A point cloud is a discrete dataset of points representing the spatial distribution and surface characteristics of an object in a specific coordinate system. Point clouds contain rich measurement information, including the target object's three-dimensional coordinates (XYZ), color (RGB), and laser reflectance intensity. They also possess properties of being disordered, permutation-invariant, and rotation-invariant. Because of these characteristics, with the development of computer and sensor technologies, the application scope of point cloud data processing technology has been continuously expanding, extending into fields such as autonomous driving, precision measurement, geographic mapping, cultural relic preservation, and medicine. Depending on the measurement scope and application, point cloud acquisition methods vary, primarily including contact scanners, optical scanners, and stereo vision. Additionally, point clouds can also be obtained from 3D models.

[0004] Point cloud registration is a crucial step in point cloud data processing, primarily addressing the spatial positional differences between the aligned point cloud and the target point cloud. It can be divided into two stages: coarse registration and fine registration. Coarse registration involves registering the point clouds when their relative poses are completely unknown, providing a good initial value for fine registration. Fine registration, based on the results of coarse registration, typically reduces the spatial positional differences between point clouds through iterative calculations. By improving registration accuracy, the positional results of point cloud objects can be better represented, thereby improving the performance and accuracy of subsequent tasks.

[0005] However, current point cloud registration methods not only have large registration errors but also take a long time to register, making them increasingly unable to meet the needs of various fields for point cloud registration. Summary of the Invention

[0006] To address the technical problems mentioned above, this invention provides a point cloud registration method and system based on a multi-feature point set, which effectively improves the accuracy and efficiency of point cloud matching and reduces registration time.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The first aspect of the present invention provides a point cloud registration method based on a multi-feature point set.

[0009] A point cloud registration method based on a multi-feature point set includes:

[0010] Acquire aligned point cloud data and target point cloud data, and preprocess the aligned point cloud data and target point cloud data;

[0011] Extract the vertices and center points of the OBB bounding boxes from the aligned point cloud data, as well as the vertices and center points of the OBB bounding boxes from the target point cloud data. Based on the aligned point cloud data, construct a hybrid KD tree, extract the FPFH features of each point, construct the FPFH matrix, calculate the mean and standard deviation of the FPFH matrix, and adjust the parameters of the ISS algorithm based on the mean and standard deviation. Use the optimized ISS algorithm to extract the ISS feature points from the aligned point cloud data. Use the same method to obtain the ISS feature points from the target point cloud data. Extract the geometric centroids of the aligned point cloud data and the target point cloud data respectively.

[0012] Select at least one of the following data from the OBB bounding box of the aligned point cloud data, the ISS feature points of the aligned point cloud data, and the geometric centroid of the aligned point cloud data to construct a first feature point set; select at least one of the following data from the OBB bounding box of the target point cloud data, the ISS feature points of the target point cloud data, and the geometric centroid of the target point cloud data to construct a second feature point set.

[0013] Based on the first feature point set, the second feature point set, the aligned point cloud data, and the target point cloud data, coarse registration and fine registration are performed sequentially.

[0014] Furthermore, the preprocessing process includes performing Gaussian distribution statistical noise reduction on both the aligned point cloud data and the target point cloud data.

[0015] Furthermore, the preprocessing process also includes voxel downsampling of both the aligned point cloud data and the target point cloud data.

[0016] Furthermore, after calculating the mean and standard deviation of the FPFH matrix, the process further includes:

[0017] If the mean is greater than or equal to 1, then read the number of decimal places in the mean and record it as n1, and update the mean to μ' = μ / 10. n1 If the mean is less than 1, then the number of digits between the decimal point and the first non-zero digit of the mean is recorded as m1, and the mean is updated to μ' = μ / 10. -m1 ;

[0018] If the standard deviation σ is greater than or equal to 1, then the number of decimal places before the decimal point of the mean is recorded as n2, and the standard deviation is updated to σ' = σ / 10. n2If the standard deviation is less than 1, the number of decimal places between the decimal point and the first non-zero value of the mean is recorded as m2, and the standard deviation is updated to σ' = σ / 10. -m2 ;

[0019] The updated mean and standard deviation are obtained.

[0020] Furthermore, the process of adjusting the parameters of the ISS algorithm based on the mean and standard deviation includes:

[0021] Determine γ 12 Is the value of μ'+σ' greater than 1?

[0022] If γ 12 If the result is greater than 1, then the larger of μ' and σ' is determined.

[0023] When μ' is greater than σ', substitute it into the formula μ' = μ' / 10 and iterate;

[0024] When μ' is less than or equal to σ', substitute the formula σ' = σ' / 10 for iteration;

[0025] Substitute the iterated data back into the formula γ 12 After =μ'+σ', determine γ 12 Is it greater than 1? If γ 12 If the value is not greater than 1, no change will be made;

[0026] According to γ 12 =μ'+σ' and γ 23 =σ', which gives the parameters of the adjusted ISS algorithm;

[0027] Where, γ 12 and γ 23 These are all parameters of the ISS algorithm.

[0028] Furthermore, the coarse registration process includes:

[0029] Both the first feature point set and the second feature point set are transformed into the first feature point matrix and the second feature point matrix;

[0030] Calculate the mean and covariance matrix of the first feature point matrix, and the mean and covariance matrix of the second feature point matrix;

[0031] The singular value decomposition algorithm is used to obtain the eigenvalues ​​and eigenvectors of the covariance matrix, and the eigenvectors and centroids are returned.

[0032] The registration error is calculated iteratively based on different principal axis directions, and the transformation matrix with the smallest error is finally selected as the final point cloud registration result.

[0033] Furthermore, the fine registration process includes: based on the aligned point cloud data and the target point cloud data, using the G-ICP algorithm and combining the transformation matrix with the minimum error determined by the coarse registration, fine registration is performed to obtain the registration results of the aligned point cloud data, the target point cloud data, the first feature point set, and the second feature point set.

[0034] A second aspect of the present invention provides a point cloud registration system based on a multi-feature point set.

[0035] A point cloud registration system based on a multi-feature point set includes:

[0036] The data acquisition module is configured to acquire aligned point cloud data and target point cloud data, and to preprocess the aligned point cloud data and target point cloud data.

[0037] The feature point extraction module is configured to: extract the vertices and center points of the OBB bounding boxes of the aligned point cloud data and the target point cloud data; construct a hybrid KD tree based on the aligned point cloud data, extract the FPFH features of each point, construct the FPFH matrix, calculate the mean and standard deviation of the FPFH matrix, and adjust the parameters of the ISS algorithm based on the mean and standard deviation; extract the ISS feature points of the aligned point cloud data using the optimized ISS algorithm; obtain the ISS feature points of the target point cloud data using the same method; and extract the geometric centroids of the aligned point cloud data and the target point cloud data respectively.

[0038] The point set construction module is configured to: select at least one of the following data from the OBB bounding box of the aligned point cloud data, the ISS feature points of the aligned point cloud data, and the geometric centroid of the aligned point cloud data to construct a first feature point set; and select at least one of the following data from the OBB bounding box of the target point cloud data, the ISS feature points of the target point cloud data, and the geometric centroid of the target point cloud data to construct a second feature point set.

[0039] The registration module is configured to perform coarse registration and fine registration sequentially based on the first feature point set, the second feature point set, the aligned point cloud data, and the target point cloud data.

[0040] A third aspect of the present invention provides a computer-readable storage medium.

[0041] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the point cloud registration method based on a multi-feature point set as described in the first aspect above.

[0042] A fourth aspect of the present invention provides a computer device.

[0043] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the point cloud registration method based on a multi-feature point set as described in the first aspect above.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] This invention employs a statistical denoising and voxel downsampling algorithm based on Gaussian distribution to effectively remove outlier noise, reduce data volume, and improve processing efficiency while preserving the original cloud structure characteristics.

[0046] This invention establishes an adaptive parameter model for extracting Intrinsic Shape Signature (ISS) feature points. Based on the Fast Point Feature Histogram (FPFH) features and average density of the point cloud, it adaptively sets the threshold and neighborhood radius in the algorithm, providing effective ISS feature points for the subsequent registration process in this invention. This invention employs a method of coarse registration with multiple feature point sets and fine registration of the overall point cloud. By extracting OBB (Oriented Bounding Bix) feature points, point cloud centroids, and ISS feature points, and selecting multiple feature points to form a point set for coarse registration, it provides an initial registration matrix for the overall fine registration. This reduces the amount of data processing required for registration, improves registration convergence and efficiency, and reduces registration errors. Attached Figure Description

[0047] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0048] Figure 1 This is a schematic diagram of the point cloud registration method based on multi-feature point set extraction according to the present invention;

[0049] Figure 2 This is a schematic diagram of the improved ISS feature point extraction algorithm with adaptive parameters according to the present invention.

[0050] Figure 3 Align the initial position maps of the point cloud and the original point cloud for the Stanford open-source dataset Bunny model;

[0051] Figure 4 Comparison of registration results for the Bunny model on the Stanford open-source dataset: Figure 4 (a) shows the initial position of the multi-feature point set, (b) shows the coarse registration result of the multi-feature point set, and (c) shows the fine registration result of the overall point cloud.

[0052] Figure 5 Align the initial position maps of the point cloud and the original point cloud for the Stanford open-source dataset Monkey model;

[0053] Figure 6 Comparison of registration results for the Monkey model on the Stanford open-source dataset: Figure 6 (a) shows the initial position of the multi-feature point set, (b) shows the coarse registration result of the multi-feature point set, and (c) shows the fine registration result of the overall point cloud. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0055] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0056] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0057] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and systems according to various embodiments of this disclosure. It should be noted that each block in a flowchart or block diagram may represent a module, segment, or portion of code, which may include one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutively represented blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, may be implemented using a dedicated hardware-based system that performs the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0058] Example 1

[0059] This embodiment provides a point cloud registration method based on a multi-feature point set. This embodiment uses the application of this method to a server as an example for illustration. It is understood that this method can also be applied to terminals, and can also be applied to systems including terminals, servers, and other components, and can be implemented through interaction between the terminal and the server. The server can be a standalone physical server, a server cluster composed of multiple physical servers, or a distributed system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communication, middleware services, domain name services, CDN security services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein. In this embodiment, the method includes the following steps:

[0060] Acquire aligned point cloud data and target point cloud data, and preprocess the aligned point cloud data and target point cloud data;

[0061] Extract the vertices and center points of the OBB bounding boxes from the aligned point cloud data, as well as the vertices and center points of the OBB bounding boxes from the target point cloud data. Based on the aligned point cloud data, construct a hybrid KD tree, extract the FPFH features of each point, construct the FPFH matrix, calculate the mean and standard deviation of the FPFH matrix, and adjust the parameters of the ISS algorithm based on the mean and standard deviation. Use the optimized ISS algorithm to extract the ISS feature points from the aligned point cloud data. Use the same method to obtain the ISS feature points from the target point cloud data. Extract the geometric centroids of the aligned point cloud data and the target point cloud data respectively.

[0062] Select at least one of the following data from the OBB bounding box of the aligned point cloud data, the ISS feature points of the aligned point cloud data, and the geometric centroid of the aligned point cloud data to construct a first feature point set; select at least one of the following data from the OBB bounding box of the target point cloud data, the ISS feature points of the target point cloud data, and the geometric centroid of the target point cloud data to construct a second feature point set.

[0063] Based on the first feature point set, the second feature point set, the aligned point cloud data, and the target point cloud data, coarse registration and fine registration are performed sequentially.

[0064] The technical solution of this embodiment will be described in detail below with reference to the accompanying drawings, such as... Figure 1 As shown:

[0065] Step 1: Point cloud data reading and preprocessing.

[0066] Step 1.1: Read the aligned point cloud Sn and target point cloud T n The files can be read in formats including but not limited to .pcd, .ply, and .obj. The aligned point cloud is a point cloud acquired via laser or other methods; the target point cloud is a point cloud generated from a theoretical model.

[0067] Step 1.2: Process the point cloud S respectively n and T n Perform statistical denoising using a Gaussian distribution. Set the number of neighboring points within the average distance and a threshold based on the standard deviation of the average distance in the point cloud. Identify and remove outliers in the point cloud to obtain point cloud S and target point cloud T.

[0068] Step 1.3: Perform voxel downsampling on point clouds S and T respectively. Set the voxel size by dividing the point cloud data into a regular voxel grid, and then retaining only one point in each voxel to obtain the denoised and simplified point cloud S. d and T d .

[0069] Step 2: Construct a multi-feature point set.

[0070] Step 2.1: Calculate the point cloud S respectively d and T d Find the OBB bounding box and extract the vertices OBBb of the OBB bounding box. S ,OBBb T and center point OBBg S ,OBBg T .

[0071] Specifically, principal component analysis is used to obtain the three principal directions and covariance matrix of the point cloud.

[0072] Find the eigenvalues ​​and eigenvectors of the covariance matrix; the eigenvectors are the principal directions.

[0073] Using the principal direction and the centroid, the point cloud is translated to the origin, and the principal direction is made to coincide with the coordinate system direction to obtain the transformed point cloud.

[0074] Based on the transformed point cloud coordinates, the maximum and minimum values ​​in the x, y, and z directions are obtained, which are the vertices of the bounding box of the transformed point cloud, and thus the center point is obtained.

[0075] By performing an inverse coordinate transformation on the transformed bounding box vertices and center point, the bounding box vertices and center point of the original point cloud are obtained.

[0076] Step 2.2: Calculate the point cloud S using the adaptive parameter model designed in this invention. d and T d ISS feature points, such as Figure 2The flowchart of the improved ISS feature point extraction algorithm with adaptive parameters is shown.

[0077] Step 2.2.1: Construct a hybrid KD-tree for the point cloud, establishing K-nearest neighbor and radius neighborhood relationships to obtain the FPFH features for each point. Use the KD-tree to perform a nearest neighbor search for each point, obtaining the distance from each point to its nearest neighbor. Calculate the average density, denoted as D, by averaging these distances. S and D T .

[0078] Step 2.2.2: Extract the FPFH matrix and calculate its mean μ and standard deviation σ. If the mean μ is greater than or equal to 1, read the number of decimal places before the decimal point of the mean and record it as n1, then output μ' = μ / 10. n1 If the mean is less than 1, then read the number of digits between the decimal point and the first non-zero digit of the mean and record it as m1, then output μ' = μ / 10. -m1 If the standard deviation σ is greater than or equal to 1, then read the number of decimal places before the decimal point of the mean and record it as n2, and output σ' = σ / 10. n2 If the standard deviation is less than 1, record the number of decimal places between the decimal point of the mean and the first non-zero value as m2, and output σ' = σ / 10. -m2 Record the final output μ' and σ' as parameters for subsequent algorithms.

[0079] Step 2.2.3: Establish an adaptive parameter model for ISS feature point calculation. Determine γ. 12 Is the value of μ'+σ' greater than 1? If γ 12 If the result is greater than 1, then the larger of μ' and σ' is determined. When μ' is greater than σ', the formula μ' = μ' / 10 is used for iteration; when μ' is less than or equal to σ', the formula σ' = σ' / 10 is used for iteration. The iterated data is then substituted back into the formula γ. 12 =μ'+σ' then determine γ 12 Is it greater than 1? If γ 12 If the value is not greater than 1, no change will be made.

[0080] Step 2.2.4: Set the parameters of the ISS algorithm using the following formula, γ 12 =μ'+σ',γ 23 =σ', Salient radius = 3*D i (i is S or T). Obtain the point cloud S respectively. d and T d Feature points ISS S and ISS T .

[0081] Step 2.3: Use the get_center function in the Open3D library to extract the point cloud S.d and T d Geometric centroid GEO S and GEO T .

[0082] Step 3: Perform coarse registration on the multi-feature point set.

[0083] Step 3.1: Based on point cloud models with different geometric structures, select one or more feature points from the above four types of feature points and combine them to construct a feature point set. Furthermore, the point cloud S needs to be... d and T d Selecting the same combination of feature points, each is denoted as the first feature point set K. S Second feature point set K T .

[0084] Step 3.2: Using a coarse registration algorithm based on principal component analysis, the two sets of feature points K extracted in Step 3.1 are matched. S and K T Convert to point cloud KP S and KP T Then, PCA coarse registration is performed.

[0085] Specifically, calculate point cloud KP S and KP T The mean and covariance matrices are obtained, and then eigenvalues ​​and eigenvectors of the covariance matrix are obtained using Singular Value Decomposition (SVD). The eigenvectors and centroids are returned. The registration error is iteratively calculated based on different principal axis directions, and the transformation matrix with the smallest error is selected as the final point cloud registration result, obtaining the initial transformation matrix T. PCA .

[0086] Step 4: For the target point cloud S d Alignment point cloud T d Perform fine registration. Read the initial transformation matrix T obtained in step 3. PCA Fine registration was performed using the G-ICP (Generalized Iterative ClosestPoint) algorithm optimized based on the L2LOSS robust loss function. The accuracy of the registration results was evaluated by calculating the root mean square error (RMSE) of the inliers after registration and the set of correspondences. The smaller the RMSE value, the smaller the difference between corresponding points, and the more accurate the point cloud registration result; the larger the number of corresponding set points, the more successful matching pairs in the point cloud, and the higher the matching quality of the registration.

[0087] Specifically, the G-ICP (Generalized Iterative ClosestPoint) algorithm based on L2LOSS robust loss function optimization involves adding a probabilistic model to the cost function of the ICP algorithm. The L2LOSS loss function is as follows:

[0088] L(Y,f(x))=(Yf(x)) 2

[0089] Where Y represents the target point cloud and f(x) represents the aligned point cloud.

[0090] Visualization in Open3D is used to visualize point clouds, showing two sets of point cloud S. d and T d and feature point cloud KP S and KP T Initial position and registration results.

[0091] The technical effectiveness of the point cloud registration method based on multi-feature point set extraction proposed in this invention is demonstrated by selecting commonly used open-source point cloud models in this field. The Stanford University database Bunny model and Monkey model are used for verification.

[0092] As can be seen from Tables 1 and 2, the present invention selects different feature points for combination, such as bounding box center point + point cloud geometric centroid point + ISS feature point (OBBg + GEO + ISS), point cloud geometric centroid point + ISS feature point (GEO + ISS), and ISS feature point (ISS).

[0093] Table 1 Comparison of registration results for the Bunny model

[0094]

[0095] Table 2 Comparison of Registration Results of Monkey Model

[0096]

[0097] The algorithm of this invention is compared with traditional registration algorithms. Figure 3 Align the initial positions of the point cloud and the original point cloud for the Bunny model. Figure 4 This is a comparison chart of the registration results under the Bunny model. Figure 4 (a) shows the initial position of the multi-feature point set, (b) shows the coarse registration result of the multi-feature point set, and (c) shows the fine registration result of the overall point cloud. Figure 5 Align the initial positions of the point cloud and the original point cloud for the Monkey model. Figure 6 This is a comparison chart of the registration results under the Monkey model. Figure 6In the diagram, (a) shows the initial position of the multi-feature point set, (b) shows the coarse registration result of the multi-feature point set, and (c) shows the fine registration result of the overall point cloud. As can be seen from the above figures, the adaptive parameter ISS feature point improvement algorithm in this invention can effectively extract feature points from the point cloud and achieves good registration results after coarse registration.

[0098] For the Bunny model, the registration of different feature point sets in this invention improves the efficiency by an average of approximately 38.22%, 40.49%, and 17.13% compared to the traditional G-ICP point cloud registration algorithm, respectively. For the Monkey model, the registration of different feature point sets in this invention improves the efficiency by an average of several tens of orders of magnitude compared to the traditional G-ICP point cloud registration algorithm. Simultaneously, the registration time required for both models is reduced compared to traditional registration methods.

[0099] This invention improves the ISS feature point extraction algorithm by establishing an adaptive parameter model, and constructs a multi-feature point set by combining the OBB bounding box and the geometric centroid of the point cloud. Then, the initial transformation matrix is ​​obtained by coarse registration of this point set, which improves the registration accuracy and reduces the registration time, thus having high registration efficiency.

[0100] Example 2

[0101] This embodiment provides a point cloud registration system based on a multi-feature point set.

[0102] A point cloud registration system based on a multi-feature point set includes:

[0103] The data acquisition module is configured to acquire aligned point cloud data and target point cloud data, and to preprocess the aligned point cloud data and target point cloud data.

[0104] The feature point extraction module is configured to: extract the vertices and center points of the OBB bounding boxes of the aligned point cloud data and the target point cloud data; construct a hybrid KD tree based on the aligned point cloud data, extract the FPFH features of each point, construct the FPFH matrix, calculate the mean and standard deviation of the FPFH matrix, and adjust the parameters of the ISS algorithm based on the mean and standard deviation; extract the ISS feature points of the aligned point cloud data using the optimized ISS algorithm; obtain the ISS feature points of the target point cloud data using the same method; and extract the geometric centroids of the aligned point cloud data and the target point cloud data respectively.

[0105] The point set construction module is configured to: select at least one of the following data from the OBB bounding box of the aligned point cloud data, the ISS feature points of the aligned point cloud data, and the geometric centroid of the aligned point cloud data to construct a first feature point set; and select at least one of the following data from the OBB bounding box of the target point cloud data, the ISS feature points of the target point cloud data, and the geometric centroid of the target point cloud data to construct a second feature point set.

[0106] The registration module is configured to perform coarse registration and fine registration sequentially based on the first feature point set, the second feature point set, the aligned point cloud data, and the target point cloud data.

[0107] It should be noted that the data acquisition module, feature point extraction module, point set construction module, and registration module described above are the same examples and application scenarios implemented in the steps of Embodiment 1, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0108] Example 3

[0109] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the point cloud registration method based on a multi-feature point set as described in Embodiment 1 above.

[0110] Example 4

[0111] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the point cloud registration method based on a multi-feature point set as described in Embodiment 1 above.

[0112] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0113] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0114] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0115] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0116] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0117] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A point cloud registration method based on a multi-feature point set, characterized in that, include: Acquire aligned point cloud data and target point cloud data, and preprocess the aligned point cloud data and target point cloud data; Extract the vertices and center points of the OBB bounding box of the aligned point cloud data, as well as the vertices and center points of the OBB bounding box of the target point cloud data; Based on aligned point cloud data, a hybrid KD tree is constructed, FPFH features of each point are extracted, an FPFH matrix is ​​constructed, and the mean and standard deviation of the FPFH matrix are calculated. Based on the mean and standard deviation, the parameters of the ISS algorithm are adjusted. The optimized ISS algorithm is then used to extract ISS feature points from the aligned point cloud data. The same method is used to obtain the ISS feature points of the target point cloud data. Extract the geometric centroids of the aligned point cloud data and the target point cloud data respectively; Select at least one of the following data from the OBB bounding box of the aligned point cloud data, the ISS feature points of the aligned point cloud data, and the geometric centroid of the aligned point cloud data to construct the first feature point set; Select at least one of the following data from the target point cloud data: the vertices and center points of the OBB bounding box, the ISS feature points of the target point cloud data, and the geometric centroid points of the target point cloud data, to construct a second feature point set; Based on the first feature point set, the second feature point set, the aligned point cloud data, and the target point cloud data, coarse registration and fine registration are performed sequentially.

2. The point cloud registration method based on a multi-feature point set according to claim 1, characterized in that, The preprocessing process includes: performing Gaussian distribution statistical noise reduction on both the aligned point cloud data and the target point cloud data.

3. The point cloud registration method based on a multi-feature point set according to claim 1, characterized in that, The preprocessing process also includes voxel downsampling of both the aligned point cloud data and the target point cloud data.

4. The point cloud registration method based on a multi-feature point set according to claim 1, characterized in that, The calculation of the mean and standard deviation of the FPFH matrix also includes: If the mean is greater than or equal to 1, then read the number of decimal places in the mean and record it as n1, and update the mean to μ' = μ / 10. n1 If the mean is less than 1, then the number of digits between the decimal point and the first non-zero digit of the mean is recorded as m1, and the mean is updated to μ' = μ / 10. -m1 ; If the standard deviation σ is greater than or equal to 1, then the number of decimal places before the decimal point of the mean is recorded as n2, and the standard deviation is updated to σ' = σ / 10. n2 If the standard deviation is less than 1, the number of decimal places between the decimal point and the first non-zero value of the mean is recorded as m2, and the standard deviation is updated to σ' = σ / 10. -m2 ; The updated mean and standard deviation are obtained.

5. The point cloud registration method based on a multi-feature point set according to claim 4, characterized in that, The process of adjusting the parameters of the ISS algorithm based on the mean and standard deviation includes: Determine γ 12 Is the value of μ'+σ' greater than 1? If γ 12 If the result is greater than 1, then the larger of μ' and σ' is determined. When μ' is greater than σ', substitute it into the formula μ' = μ' / 10 and iterate; When μ' is less than or equal to σ', substitute the formula σ' = σ' / 10 for iteration; Substitute the iterated data back into the formula γ 12 After =μ'+σ', determine γ 12 Is it greater than 1? If γ 12 If the value is not greater than 1, no change will be made; According to γ 12 =μ'+σ' and γ 23 =σ', which gives the parameters of the adjusted ISS algorithm; Where, γ 12 and γ 23 These are all parameters of the ISS algorithm.

6. The point cloud registration method based on a multi-feature point set according to claim 1, characterized in that, The coarse registration process includes: Both the first feature point set and the second feature point set are transformed into the first feature point matrix and the second feature point matrix; Calculate the mean and covariance matrix of the first feature point matrix, and the mean and covariance matrix of the second feature point matrix; The singular value decomposition algorithm is used to obtain the eigenvalues ​​and eigenvectors of the covariance matrix, and the eigenvectors and centroids are returned. The registration error is calculated iteratively based on different principal axis directions, and the transformation matrix with the smallest error is finally selected as the final point cloud registration result.

7. The point cloud registration method based on a multi-feature point set according to claim 1, characterized in that, The process of fine registration includes: based on the aligned point cloud data and the target point cloud data, using the G-ICP algorithm and combining the transformation matrix with the minimum error determined by the coarse registration, fine registration is performed to obtain the registration results of the aligned point cloud data, the target point cloud data, the first feature point set, and the second feature point set.

8. A point cloud registration system based on a multi-feature point set, characterized in that, include: The data acquisition module is configured to acquire aligned point cloud data and target point cloud data, and to preprocess the aligned point cloud data and target point cloud data. The feature point extraction module is configured to extract the vertices and center points of the OBB bounding box of the aligned point cloud data, as well as the vertices and center points of the OBB bounding box of the target point cloud data. Based on aligned point cloud data, a hybrid KD tree is constructed, FPFH features of each point are extracted, an FPFH matrix is ​​constructed, and the mean and standard deviation of the FPFH matrix are calculated. Based on the mean and standard deviation, the parameters of the ISS algorithm are adjusted. The optimized ISS algorithm is then used to extract ISS feature points from the aligned point cloud data. The same method is used to obtain the ISS feature points of the target point cloud data. Extract the geometric centroids of the aligned point cloud data and the target point cloud data respectively; The point set construction module is configured to: select at least one of the following data: the vertices and center points of the OBB bounding box of the aligned point cloud data, the ISS feature points of the aligned point cloud data, and the geometric centroid points of the aligned point cloud data, to construct a first feature point set; Select at least one of the following data from the target point cloud data: the vertices and center points of the OBB bounding box, the ISS feature points of the target point cloud data, and the geometric centroid points of the target point cloud data, to construct a second feature point set; The registration module is configured to perform coarse registration and fine registration sequentially based on the first feature point set, the second feature point set, the aligned point cloud data, and the target point cloud data.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the point cloud registration method based on a multi-feature point set as described in any one of claims 1-7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the point cloud registration method based on a multi-feature point set as described in any one of claims 1-7.

Images