CN112102376B - Multi-view cloud registration method, device and storage medium of hybrid sparse ICP - Google Patents
Multi-view cloud registration method, device and storage medium of hybrid sparse ICP Download PDFInfo
- Publication number
- CN112102376B CN112102376B CN202010773290.9A CN202010773290A CN112102376B CN 112102376 B CN112102376 B CN 112102376B CN 202010773290 A CN202010773290 A CN 202010773290A CN 112102376 B CN112102376 B CN 112102376B
- Authority
- CN
- China
- Prior art keywords
- point cloud
- target point
- registration
- sparse
- view
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Abstract
The invention discloses a multi-view cloud registration method, a device and a storage medium of mixed sparse ICP (inductively coupled plasma). Firstly, two pieces of 3D point clouds with different view angles can be obtained through a laser scanner, after the target point clouds are initialized through KD-Tree, the KD-Tree is utilized to initialize the target point clouds, the solving speed of nearest neighbor points of the target point clouds can be accelerated, then, aiming at the sparse expression characteristics of outliers and noise in multi-view cloud registration, a mixed sparse model is established, a regular term with 2 norms is added into a P norms sparse model required by multi-view cloud registration, and the high sparse and near sparse representation capability of the registration model is endowed, so that the accurate registration of two pieces of adjacent multi-view clouds is realized. The method has good anti-noise and generalization performance, and can avoid the convergence of the solving multi-view cloud registration transformation matrix to a local optimal solution. The invention is applied to the technical field of computer vision.
Description
Technical Field
The disclosure relates to the technical field of computer vision, in particular to a multi-view cloud registration method, a device and a storage medium of hybrid sparse ICP.
Background
The point cloud data registration is one of the most important research contents of computer graphics, and is also a key technology in applications such as object recognition, gesture estimation, face recognition, surface matching and the like. As a current research hotspot, three-dimensional reconstruction has been widely applied to various fields in life and entertainment, including manufacturing industry, medicine, archaeology and the like. And the point cloud data matching criterion is an indispensable link in the three-dimensional reconstruction process.
When point cloud registration is carried out, the single view angle is adopted for scanning, so that the limitations of small sensing range, shielding and the like exist; in addition, the accuracy of registration of neighboring multi-view point clouds with noisy, outliers tends to be poor.
Disclosure of Invention
The present disclosure aims to solve at least one of the above problems, and provides a multi-view cloud registration method, apparatus and storage medium for hybrid sparse ICP.
To achieve the above object, according to an aspect of the present disclosure, there is provided a multi-view cloud registration method of hybrid sparse ICP, the method comprising the steps of:
step 101, acquiring adjacent target point cloud A and target point cloud B, and carrying out normalization processing, wherein the target point cloud A and the target point cloud B are multi-viewpoint clouds;
102, initializing a target point cloud A through KD-Tree, and accelerating the solving speed of the nearest neighbor point;
step 103, substituting the target point cloud a and the target point cloud B into the constructed hybrid sparse registration model about the target point cloud a and the target point cloud B to obtain a spatial transformation matrix R and a corresponding registration point cloud B * The hybrid sparse registration model is:
wherein h=r+b+t-B Closest +lambda/. Mu.z is noise, ρ is weight, p is positive integer, R is space transformation matrix, lambda and mu are multiplier variables, B Closest The nearest neighbor point of the target point cloud B in the target point cloud A is set, and t is the iteration number;
104, calculating a target point cloud B and a registration point cloud B * Judging the magnitude relation between the root mean square error delta and the first threshold value, and if the root mean square error delta is smaller than the first threshold value, controlling to end the program and outputting a registration result; if the root mean square error delta is not less than the first threshold value, let b=b * Returning to step 102, continuing to execute the multi-view cloud registration method of the mixed sparse ICP.
Further, the manner of determining the weight ρ in the step 103 specifically includes the following steps:
initializing the multiplier variable μ, the spatial transformation matrix R, and the number of iterations t, i.e., let μ=10, r=i, t= [ 000 00] T The weight ρ is determined by the following formula:
further, the noise z is solved by an alternate multiplier method, which can be determined by the following formula:
further, the spatial transformation matrix R in the step 103 is obtained by combining singular value decomposition and matching the centroid of the point cloud, and specifically includes the following steps:
R＝SVD(B，C)；t＝mean(B Closest )-mean(R*B)；C＝B Closest +z-λ/μ。
further, the updating method of the multiplier variables λ and μ in step 103 specifically includes the following steps:
λ t+1 ＝λ t +μδ；δ＝B-B Closest ；μ＝1.2*μ。
the invention also provides a multi-view cloud registration device of the mixed sparse ICP, which is applied to the multi-view cloud registration method of the mixed sparse ICP, and comprises the following steps:
the target point cloud acquisition unit is used for acquiring adjacent target point clouds A and B;
the initialization unit is used for initializing the target point cloud A through KD-Tree to obtain an initialized target point cloud A;
a calculation unit for calculating a spatial transformation matrix R and a corresponding registration point cloud B according to the target point cloud A and the target point cloud B * ；
A judging unit for calculating a target point cloud B and an alignment point cloud B * Judging the magnitude relation between the root mean square error delta and the first threshold value, ending the program if the root mean square error delta is smaller than the first threshold value, and outputting a registration result; if the root mean square error delta is not less than the first threshold value, let b=b * Returning to step 102, continuing to execute the multi-view cloud registration method of the mixed sparse ICP.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor implements the steps of the hybrid sparse ICP multi-view cloud registration method of any of claims 1-5.
The beneficial effects of the present disclosure are: according to the multi-view cloud registration method, device and storage medium of the mixed sparse ICP, aiming at the sparse expression characteristics of outliers and noise during multi-view cloud registration, a mixed sparse model is established, a regular term of 2 norms is added into a P norms sparse model required by multi-view cloud registration, high sparse and near sparse representation capability of a registration model is endowed, and accurate registration of two adjacent multi-view clouds is achieved. The method has good anti-noise and generalization performance, and can avoid the convergence of the solving multi-view cloud registration transformation matrix to a local optimal solution.
Drawings
The above and other features of the present disclosure will become more apparent from the detailed description of the embodiments illustrated in the accompanying drawings, in which like reference numerals designate like or similar elements, and which, as will be apparent to those of ordinary skill in the art, are merely some examples of the present disclosure, from which other drawings may be made without inventive effort, wherein:
FIG. 1 is a flow chart of a multi-view cloud registration method of hybrid sparse ICP;
fig. 2 is a schematic structural diagram of a multi-view cloud registration device of a hybrid sparse ICP;
FIG. 3 is a flow chart of a hybrid sparse ICP algorithm;
FIG. 4 is a schematic diagram of a nearest neighbor search of a point cloud;
FIG. 5 is a schematic diagram illustrating initial positions of neighboring point clouds in one embodiment;
fig. 6 is a schematic diagram showing the registration result of the A, B point cloud after the application of the method according to an embodiment of fig. 5.
Detailed Description
The conception, specific structure, and technical effects produced by the present disclosure will be clearly and completely described below in connection with the embodiments and the drawings to fully understand the objects, aspects, and effects of the present disclosure. It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other.
Referring to fig. 1, 3 and 4, which are flowcharts illustrating a multi-view cloud registration method of hybrid sparse ICP according to the present disclosure, a multi-view cloud registration method of hybrid sparse ICP according to an embodiment of the present disclosure is described below.
The disclosure proposes a multi-view cloud registration method of hybrid sparse ICP, the method comprising the steps of:
step 101, acquiring adjacent target point cloud A and target point cloud B, and carrying out normalization processing to ensure uniform size, wherein the target point cloud A and the target point cloud B are multi-viewpoint clouds;
102, initializing a target point cloud A through KD-Tree, and accelerating the solving speed of the nearest neighbor point;
step 103, substituting the target point cloud a and the target point cloud B into the constructed hybrid sparse registration model about the target point cloud a and the target point cloud B to obtain a spatial transformation matrix R and a corresponding registration point cloud B * The hybrid sparse registration model is:
wherein h=r+b+t-B Closest +lambda/. Mu.z is noise, ρ is weight, p is positive integer, R is space transformation matrix, lambda and mu are multiplier variables, B Closest The nearest neighbor point of the target point cloud B in the target point cloud A is set, and t is the iteration number;
104, calculating a target point cloud B and a registration point cloud B * Judging the magnitude relation between the root mean square error delta and the first threshold value, and if the root mean square error delta is smaller than the first threshold value, controlling to end the program and outputting a registration result; if the root mean square error delta is not less than the first threshold value, let b=b * Returning to step 102, continuing to execute the multi-view cloud registration method of the mixed sparse ICP.
In the embodiment, two 3D point clouds with different view angles can be obtained through a laser scanner, after the target point clouds are initialized through a KD-Tree, the target point clouds are initialized through the KD-Tree, the solving speed of the nearest neighbor points of the target point clouds can be accelerated, and then aiming at the sparse expression characteristics of outliers and noise during multi-view cloud registration, a mixed sparse model is established, a regular term with 2 norms is added into a P-norms sparse model required by multi-view cloud registration, so that the high sparse and near sparse representation capability of the registration model is endowed, and the accurate registration of the two adjacent multi-view clouds is realized. The method has good anti-noise and generalization performance, and can avoid the convergence of the solving multi-view cloud registration transformation matrix to a local optimal solution.
Specifically, the first threshold is a condition for ending the iteration, and in this embodiment, the first threshold is set to 0.0001, which can better implement the method of the present scheme.
As a preferred embodiment of the present solution, the manner of determining the weight ρ in the step 103 specifically includes the following:
initializing the multiplier variable μ, the spatial transformation matrix R, and the number of iterations t, i.e., let μ=10, r=i, t= [ 000 00] T The weight ρ is determined by the following formula:
in this embodiment, a preferred embodiment of the initialization is given, and other reasonable initialization modes may be selected on the premise of realizing the present invention.
As a preferred embodiment of the present solution, the noise z is solved by an alternate multiplier method, which can be determined by the following formula:
in this embodiment, a preferred way of solving the noise z by the alternative multiplier method is given, which can be well adapted to example 1.
As a preferred embodiment of the present solution, the spatial transformation matrix R in the step 103 is obtained by combining singular value decomposition and matching the centroid of the point cloud, and specifically includes the following steps:
R＝SVD(B，C)；t＝mean(B Closest )-mean(R*B)；C＝B Closest +z-λ/μ。
as a preferred embodiment of the present embodiment, the updating method of the multiplier variables λ and μ in step 103 specifically includes the following steps:
λ t+1 ＝λ t +μδ；δ＝B-B Closest ；μ＝1.2*μ。
referring to fig. 2, the invention further provides a multi-view cloud registration device of the mixed sparse ICP, and the device is applied to the multi-view cloud registration method of the mixed sparse ICP, and comprises the following steps:
the target point cloud acquisition unit is used for acquiring adjacent target point clouds A and B;
the initialization unit is used for initializing the target point cloud A through KD-Tree to obtain an initialized target point cloud A;
a calculation unit for calculating a spatial transformation matrix R and a corresponding registration point cloud B according to the target point cloud A and the target point cloud B * ；
A judging unit for calculating a target point cloud B and an alignment point cloud B * Judging the magnitude relation between the root mean square error delta and the first threshold value, ending the program if the root mean square error delta is smaller than the first threshold value, and outputting a registration result; if the root mean square error delta is not less than the first threshold value, let b=b * Returning to step 102, continuing to execute the multi-view cloud registration method of the mixed sparse ICP.
After the device is applied to the multi-view cloud registration method of the mixed sparse ICP, the mixed sparse model is established according to the sparse expression characteristics of outliers and noise during multi-view cloud registration, and a regular term of 2 norms is added into a P norms sparse model required by multi-view cloud registration, so that high sparse and near sparse representation capability of the registration model is endowed, and accurate registration of two adjacent multi-view clouds is realized. The method has good anti-noise and generalization performance, and can avoid the convergence of the solving multi-view cloud registration transformation matrix to a local optimal solution. In specific application, fig. 5 shows initial conditions of the target point cloud a and the target point cloud B, and fig. 6 shows registration results of the target point cloud a and the target point cloud B after being processed by the multi-view point cloud registration method of the mixed sparse ICP.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor implements the steps of the hybrid sparse ICP multi-view cloud registration method of any of claims 1-5.
The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical modules, i.e., may be located in one place, or may be distributed over a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module. The integrated modules may be implemented in hardware or in software functional modules.
The integrated modules, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on this understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of each of the method embodiments described above. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the computer readable medium may include content that is subject to appropriate increases and decreases as required by jurisdictions in which such content is subject to legislation and patent practice, such as in certain jurisdictions in which such content is not included as electrical carrier signals and telecommunication signals.
While the present invention has been described in considerable detail and with particularity with respect to several described embodiments, it is not intended to be limited to any such detail or embodiments or any particular embodiment, but is to be construed as providing broad interpretation of such claims by reference to the appended claims in view of the prior art so as to effectively encompass the intended scope of the invention. Furthermore, the foregoing description of the invention has been presented in its embodiments contemplated by the inventors for the purpose of providing a useful description, and for the purposes of providing a non-essential modification of the invention that may not be presently contemplated, may represent an equivalent modification of the invention.
The present invention is not limited to the above embodiments, but is merely preferred embodiments of the present invention, and the present invention should be construed as being limited to the above embodiments as long as the technical effects of the present invention are achieved by the same means. Various modifications and variations are possible in the technical solution and/or in the embodiments within the scope of the invention.
Claims (6)
1. The multi-view cloud registration method of the mixed sparse ICP is characterized by comprising the following steps of:
step 101, acquiring adjacent target point cloud A and target point cloud B, and carrying out normalization processing, wherein the target point cloud A and the target point cloud B are multi-viewpoint clouds;
102, initializing a target point cloud A through KD-Tree, and accelerating the solving speed of the nearest neighbor point;
step 103, substituting the target point cloud a and the target point cloud B into the constructed hybrid sparse registration model about the target point cloud a and the target point cloud B to obtain a spatial transformation matrix R and a corresponding registration point cloud B * The hybrid sparse registration model is:
wherein h=r+b+t-B Closest +lambda/. Mu.z is noise, ρ is weight, p is positive integer, R is space transformation matrix, lambda and mu are multiplier variables, B Closest The nearest neighbor point of the target point cloud B in the target point cloud A is set, and t is the iteration number;
104, calculating a target point cloud B and a registration point cloud B * Judging the magnitude relation between the root mean square error delta and the first threshold value, and if the root mean square error delta is smaller than the first threshold value, controlling to end the program and outputting a registration result; if the root mean square error delta is not less than the first threshold value, let b=b * Returning to the step 102 to continue to execute the multi-view cloud registration method of the mixed sparse ICP;
the method for determining the weight ρ in step 103 specifically includes the following steps:
initializing the multiplier variable μ, the spatial transformation matrix R, and the number of iterations t, i.e., let μ=10, r=i, t= [ 000 00] T The weight ρ is determined by the following formula:
2. the method of multi-view cloud registration of mixed sparse ICP of claim 1 wherein the noise z is solved by an alternate multiplier method, determined by the formula:
3. the method according to claim 1, wherein the spatial transformation matrix R in the step 103 is obtained by combining singular value decomposition and matching of point cloud centroids, and specifically comprises the following steps:
R＝SVD(B，C)；t＝mean(B Closest )-mean(R*B)；C＝B Closest +z-λ/μ。
4. the multi-view cloud registration method of mixed sparse ICP of claim 1, wherein the updating method of the multiplier variables λ and μ in step 103 specifically includes the following steps:
λ t+1 ＝λ t +μδ；δ＝B-B Closest ；μ＝1.2*μ。
5. a multi-view cloud registration device of mixed sparse ICP, wherein the device applies the multi-view cloud registration method of mixed sparse ICP according to any one of claims 1-4, comprising:
the target point cloud acquisition unit is used for acquiring adjacent target point clouds A and B;
the initialization unit is used for initializing the target point cloud A through KD-Tree to obtain an initialized target point cloud A;
a calculation unit for calculating a spatial transformation matrix R and a corresponding registration point cloud B according to the target point cloud A and the target point cloud B * ；
A judging unit for calculating a target point cloud B and an alignment point cloud B * Judging the magnitude relation between the root mean square error delta and the first threshold value, ending the program if the root mean square error delta is smaller than the first threshold value, and outputting a registration result; if the root mean square error delta is not less than the first threshold value, let b=b * Returning to step 102, continuing to execute the multi-view cloud registration method of the mixed sparse ICP.
6. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed by a processor, implements the steps of the hybrid sparse ICP multi-view cloud registration method according to any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010773290.9A CN112102376B (en) | 2020-08-04 | 2020-08-04 | Multi-view cloud registration method, device and storage medium of hybrid sparse ICP |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010773290.9A CN112102376B (en) | 2020-08-04 | 2020-08-04 | Multi-view cloud registration method, device and storage medium of hybrid sparse ICP |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112102376A CN112102376A (en) | 2020-12-18 |
| CN112102376B true CN112102376B (en) | 2023-06-06 |
Family
ID=73750010
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010773290.9A Active CN112102376B (en) | 2020-08-04 | 2020-08-04 | Multi-view cloud registration method, device and storage medium of hybrid sparse ICP |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112102376B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113050074B (en) * | 2021-03-16 | 2023-08-25 | 成都信息工程大学 | Camera and laser radar calibration system and calibration method in unmanned environment perception |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107709928A (en) * | 2015-04-10 | 2018-02-16 | 欧洲原子能共同体由欧洲委员会代表 | For building figure and the method and apparatus of positioning in real time |
| CN108665491A (en) * | 2018-03-22 | 2018-10-16 | 西安电子科技大学 | A kind of quick point cloud registration method based on local reference |
| CN109029417A (en) * | 2018-05-21 | 2018-12-18 | 南京航空航天大学 | Unmanned plane SLAM method based on mixing visual odometry and multiple dimensioned map |
-
2020
- 2020-08-04 CN CN202010773290.9A patent/CN112102376B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107709928A (en) * | 2015-04-10 | 2018-02-16 | 欧洲原子能共同体由欧洲委员会代表 | For building figure and the method and apparatus of positioning in real time |
| CN108665491A (en) * | 2018-03-22 | 2018-10-16 | 西安电子科技大学 | A kind of quick point cloud registration method based on local reference |
| CN109029417A (en) * | 2018-05-21 | 2018-12-18 | 南京航空航天大学 | Unmanned plane SLAM method based on mixing visual odometry and multiple dimensioned map |
Non-Patent Citations (2)
| Title |
|---|
| 颅骨点云模型的优化配准;赵夫群;周明全;;光学精密工程(07);第244-250页 * |
| 颅骨点云模型的局部特征配准方法;赵夫群;周明全;耿国华;;中国图象图形学报(08);第108-115页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112102376A (en) | 2020-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112630766B (en) | Radar angle and distance estimation method based on tensor high-order singular value decomposition | |
| Lu et al. | Accelerated coherent point drift for automatic three-dimensional point cloud registration | |
| De-Maeztu et al. | Near real-time stereo matching using geodesic diffusion | |
| CN112102376B (en) | Multi-view cloud registration method, device and storage medium of hybrid sparse ICP | |
| CN110796691B (en) | Heterogeneous image registration method based on shape context and HOG characteristics | |
| CN111814839B (en) | Template matching method of longicorn group optimization algorithm based on self-adaptive variation | |
| CN112381862B (en) | Full-automatic registration method and device for CAD (computer-aided design) model and triangular mesh | |
| KR20220084308A (en) | Angular dictionary and direct coding modes for coding tree representations of point clouds | |
| Liang et al. | Hand-eye calibration with a new linear decomposition algorithm | |
| CN109709510A (en) | A kind of estimation method and system of coherent 2-d direction finding | |
| Leng et al. | A novel geometric flow-driven approach for quality improvement of segmented tetrahedral meshes | |
| CN113141202B (en) | MIMO space non-stationary channel estimation method based on image contour extraction | |
| CN109785372A (en) | Basis matrix robust estimation method based on soft decision optimization | |
| CN108009988A (en) | A kind of image fractal regressive method for optimizing population | |
| CN116993955A (en) | Three-dimensional model heavy topology method, device, equipment and storage medium | |
| Tan et al. | Angular resolution enhancement of real-beam scanning radar base on accelerated iterative shinkage/thresholding algorithm | |
| Celebi et al. | Color quantization of dermoscopy images using the k-means clustering algorithm | |
| Li et al. | DANI-Net: Uncalibrated Photometric Stereo by Differentiable Shadow Handling, Anisotropic Reflectance Modeling, and Neural Inverse Rendering | |
| CN112102377B (en) | Chebyshev-based ICP point cloud global optimal registration method and device | |
| Fan et al. | Multilevel complex source beam method for electromagnetic scattering problems | |
| CN111080517B (en) | Three-dimensional point cloud splicing method based on improved butterfly optimization algorithm | |
| CN112613528A (en) | Point cloud simplification method and device based on significance variation and storage medium | |
| CN113989381A (en) | Multi-camera calibration method based on improved gull algorithm | |
| Saval-Calvo et al. | Non-rigid point set registration using color and data downsampling | |
| CN109949397A (en) | A kind of depth map reconstruction method of combination laser point and average drifting |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |