CN114972459B - Point cloud registration method based on low-dimensional point cloud local feature descriptor - Google Patents
Point cloud registration method based on low-dimensional point cloud local feature descriptor Download PDFInfo
- Publication number
- CN114972459B CN114972459B CN202210611042.3A CN202210611042A CN114972459B CN 114972459 B CN114972459 B CN 114972459B CN 202210611042 A CN202210611042 A CN 202210611042A CN 114972459 B CN114972459 B CN 114972459B
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- descriptor
- neighborhood
- dimensional
- 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
- 238000000034 method Methods 0.000 title claims abstract description 20
- 239000013598 vector Substances 0.000 claims abstract description 17
- 238000005070 sampling Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 19
- 230000009466 transformation Effects 0.000 claims description 18
- 230000005484 gravity Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000004381 surface treatment Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
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
- G06T7/344—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
- G06T7/75—Determining position or orientation of objects or cameras using feature-based methods involving models
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Processing Or Creating Images (AREA)
Abstract
The application provides a point cloud registration method based on a low-dimensional point cloud local feature descriptor. Firstly, extracting key points of two point clouds through uniform sampling, and transferring the key points to a new 3D space to form a 3D descriptor through constructing a local reference coordinate system; then, encoding the neighborhood information of the key points by combining the sum of normal vector angles, the sum of curvature and the sum of distances in the neighborhood of the point cloud and forming a descriptor of a characteristic histogram of the neighborhood point; the 3D descriptor and the neighborhood point characteristic histogram descriptor together form a low-dimensional point cloud local characteristic descriptor; the proposed descriptor firstly carries out radial search on the position of the '3D' key point in a new 3D space, so that the search space of the corresponding point pair is reduced; finally, registration is carried out through a RANSAC algorithm by using a descriptor of a neighborhood point characteristic histogram. The algorithm can obtain accurate registration effect in a short time, and is suitable for the field of precise measurement with high requirements on registration effect.
Description
Technical Field
The application belongs to the field of point cloud registration, and relates to a point cloud registration method of a low-dimensional point cloud local feature descriptor.
Background
With the development of computer technology, three-dimensional modeling technology is increasingly widely applied in the fields of engineering modeling, digital city, cultural relic protection, automatic driving and the like. How to obtain the complete object surface information is one of the most fundamental problems. Due to limited viewing angles of point cloud acquisition equipment, shielding of objects, acquisition environments and the like, multi-angle and multi-batch acquisition is generally required to acquire complete surface information of the objects. The point cloud matching criterion is a process of integrating the point cloud data acquired in multiple batches and multiple angles into the same view angle. Mathematically, each point cloud data has its own coordinate system. The problem to be solved by the point cloud registration is to convert the point cloud data of different coordinate systems into the same coordinate system through coordinate system transformation. The currently proposed point cloud registration algorithm often needs to extract local features from a large amount of point cloud data, each local feature corresponds to a high-dimensional description vector, and matching pairs need to be searched from among a large amount of point pairs, so that the problems of large calculation amount, low calculation efficiency and the like are caused. Therefore, aiming at the problems, the application provides a point cloud registration method based on a low-dimensional point cloud local feature descriptor, which reduces the corresponding point search space by utilizing the key point position in the new 3D space, and then finds accurate matching by utilizing a neighborhood point feature histogram descriptor, thereby effectively improving the calculation speed and the accuracy.
Disclosure of Invention
The application aims to solve the problems of high algorithm complexity, long program running time and low registration precision of the existing mainstream three-dimensional point cloud registration algorithm, and provides a three-dimensional point cloud registration method based on a low-dimensional point cloud local feature descriptor. The application transfers the points in the neighborhood of the key points of the 3D descriptor to a new 3D space, so that the key points in the similar 3D surface are close to each other, and the search space is greatly reduced. And then constructing a neighborhood point characteristic histogram descriptor by using the sum of normal vector included angles between the key points and the neighborhood points, the sum of curvature values, the distance values and three parameters together. The 3D descriptor and the neighborhood point feature histogram descriptor together comprise a low-dimensional point cloud local feature descriptor. The descriptor has the characteristics of representativeness and high extraction speed, and improves the accuracy while greatly reducing the time consumption of point cloud registration.
The technical scheme of the application is as follows: the point cloud registration algorithm based on the low-dimensional point cloud local feature descriptor comprises the following basic steps:
step 1: acquiring point cloud data and calculating a point cloud data normal vector;
step 2: calculating the gravity center through voxel filtering, and traversing the nearest neighbor point of the gravity center as a key point;
step 3: the keypoints are transferred to a new 3D space whose new 3D coordinates are stored in the first three dimensions of the descriptor. The front three-dimensional descriptor is a 3D descriptor. Radial searches within the key point radius are set in the new 3D space, retrieving a list of possible matches.
Step 4: and constructing the three-dimensional feature descriptor by calculating normal vectors, curvatures and distances between the key points and the points in the radius neighborhood. The post three-dimensional descriptor is a neighborhood point feature histogram descriptor.
Step 5: radial nearest neighbor searches are performed using KD-Tree to find exact matches for the considered source keypoints from the list of possible matches generated above using a sample consistency algorithm by using euclidean metrics to find exact nearest neighbors in the post three-dimensional descriptor space.
The point clouds in the point cloud registration are divided into a source point cloud and a target point cloud. The source point cloud and the target point cloud are integrated in pairs or are acquired by a laser triangulation profiler and are used for small-sized workpiece point cloud or outdoor large-sized point cloud data.
The step 2 specifically comprises the following steps:
the minimum bounding box surrounding the whole point cloud is equally divided into a plurality of identical small cubes, the gravity center of each non-empty cube is calculated, and the method of searching the adjacent point of the gravity center to replace the small cube is adopted, so that downsampling is realized, and original characteristic information is maintained.
Step 2.1: establishing KD-Tree for the point cloud, counting the total number N of points in the point cloud, calculating the minimum bounding box of the point cloud and the volume V thereof, and setting L x 、L y 、L z The lengths of the bounding boxes in the x, y and z directions are respectively; the KD-Tree is fully called as K-dimension Tree, is used as a KNN retrieval data structure and is used for storing a K-Dimensional space vector point set and carrying out quick retrieval;
step 2.2: setting the side length of the small cube asThe point cloud is divided into n=l×w×h microcubes; wherein: s is a scaling factor, α is a scaling factor for adjusting the side length according to the number of point clouds, l=ceil (L x /L),w＝ceil(L y /L),h＝ceil(L z L), ceil is an upward rounding function;
step 2.3: calculating the gravity center of each non-empty small cube as p 0 (x 0 ,y 0 ,z 0 ) Whereinx i 、y i 、z i X, y, z axis coordinate values for the point cloud in each non-empty small cube.
Step 2.4: acquiring two groups of data of a source point cloud P and a target point cloud Q, and acquiring a key point set P by using a voxel filter i 、Q i Traversing the source point cloud P and the target point cloud Q by utilizing KD-Tree, and searching the point set P i 、Q i Nearest neighbor point of each point in the list, forming a new key point set P s 、Q t 。
The step 3 specifically comprises the following steps:
step 3.1: calculate key point K i Surface radius r neighborhood space ik Inner centroid coordinates mean pt ；
Step 3.2: by Surface treatment ik Subtracting the centroid coordinate mean from all the point coordinates in (a) pt Effectively generates a new surface
Step 3.3: let key point K i Subtracting centroid coordinates mean pt Obtaining
Step 3.4: we follow from a new surfaceThe correction covariance matrix COV is calculated as follows:
wherein: for readability we use q m Represented in a 3D curved surfaceFor the point in (a)q representsPoints d in (a) m ＝||q m -q|| 2 。
To create a unique local reference frame and eliminate sign ambiguity, the local x-axis and z-axis directions are oriented in most directions of the vector they represent, and finally the local y-axis is obtained by the outer product of z and x, which is y=zx.
Constructing a local reference coordinate system, and describing the following formula:
[RF] 3×3 ＝[x T y T z T ] T
then willConverting to a new 3D space, coding key point coordinates into 3D descriptors and setting radius threshold values to search a list of possible matches, wherein the following formula is as follows:
wherein [ K ] iRF ] 3×1 The values of the respective dimensions of (a) constitute a front three-dimensional descriptor of the low-dimensional point cloud local feature descriptor, the front three-dimensional descriptor being a 3D descriptor.
The step 4 specifically comprises the following steps:
step 4.1: calculate key point K i Normal vectorWith neighborhood space Surface ik Interior normal vector->Included angleThe method comprises the following steps:
wherein,,the value range is [0, pi ]]。
Key point K i Phasor methodWith neighborhood space Surface ik Interior normal vector->And the included angles are summed, and the sum is as follows:
and 4.2, calculating the eigenvector of each eigenvalue and the eigenvalue of the neighborhood covariance matrix, wherein the eigenvalue is represented by the following formula:
COV·v j ＝λ j ·v j
where j ε {1,2,3}, λ j Is the eigenvalue of the covariance matrix, v j Is a relevant feature vector.
Calculate key point K i Curvature and neighborhood space ir The interior point curvatures H are summed as follows:
wherein lambda is 1 Is lambda j Is the minimum feature value of the model.
Step 4.3: calculate key point K i With neighborhood space Surface ik The sum of the distance values of the inner points is as follows:
wherein K is ik For K i Points in the neighborhood of the radius, i is the index of the key point, and k is the index of the point in the neighborhood.
And 4.4, constructing a neighborhood point characteristic histogram descriptor according to the three values in the steps 4.1, 4.2 and 4.3.
The step 5 specifically comprises the following steps:
step 5.1: first for source keypoints K in the new 3D space iRFsource Find distance threshold T d Within the most recent target keypoint list K iRFtarget ；
Step 5.2: and (3) finding out an accurate corresponding point in the neighborhood point characteristic histogram descriptor by calculating the Euclidean distance to generate a matching point pair C.
Step 5.3: for the corresponding characteristic point pairs, a singular value decomposition algorithm is used for obtaining a three-dimensional rigid transformation matrix T from a source point cloud P to a target point cloud Q i ；
Step 5.4: from estimating a three-dimensional rigid transformation matrix T i Respectively carrying out coordinate conversion on each pair of corresponding feature points, judging whether each pair of corresponding feature points belongs to internal points according to Euclidean distance between the corresponding feature points after coordinate conversion, and obtaining the number of the internal points under the estimated three-dimensional rigid transformation matrix, wherein the number of the internal points is described as follows;
where k=1, 2,.. c ，N c The distance threshold is epsilon, which is the logarithm of the corresponding feature point in the feature matching point pair C.
Step 5.5: continuously repeating the step 5.4 to obtain the number of internal points of each pair of corresponding characteristic point estimation three-dimensional rigid transformation matrixes, and determining the estimation three-dimensional rigid transformation matrix with the maximum number of the estimated internal points as an initial three-dimensional rigid transformation matrix T * And carrying out coordinate transformation on the source point cloud, wherein the following formula is described:
whereby the matrix T is initially transformed * Registering the source point cloud to the target point cloud to obtain the initial position of the source point cloud to the target point cloud.
Compared with the prior art, the application has the following advantages:
(1) The present algorithm greatly reduces the matching point pair search space by using the key point coordinate locations in the new 3D space.
(2) Is easy to combine with various advanced key point detectors;
(3) The neighborhood point characteristic histogram descriptor is used for further determining the matching point pairs, so that the local characteristic calculation dimension is reduced, the calculation speed is greatly increased, and the registration efficiency is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings:
fig. 1 is a flow chart of a point cloud registration algorithm designed by the application.
Fig. 2 is a schematic diagram of a low-dimensional point cloud local feature sub-calculation according to the present application.
FIG. 3 is a schematic representation of a neighborhood point feature histogram descriptor calculation.
Fig. 4 is an initial pose diagram of a source point cloud and a target point cloud.
Fig. 5 is the registration result after the RANSAC random sample consensus algorithm.
Detailed Description
The following describes the embodiments of the present application further with reference to the drawings. And under the Windows 10 operating system, visual Studio2019 is selected as a programming tool, and registration processing is carried out on the multi-view point cloud data acquired by the three-dimensional measuring equipment. In the example, a 'deep vision intelligent' SR7080 scanner is adopted to acquire the point cloud data of the camera module, and a relatively accurate registration result is finally obtained. The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
figure one is an algorithm flow chart of the entire process of the present application.
In the three-dimensional point cloud post-processing, point clouds with different view angles are required to be registered, and the problems of low feature identification, large descriptor dimension, long registration time and the like in the existing automatic registration algorithm based on feature extraction and description are solved. The application provides a point cloud registration algorithm based on a point cloud local feature descriptor. The point cloud is turned into a new 3D space by constructing a local reference coordinate system, and key points similar to a curved surface are close to each other, so that the search space of a matching pair is greatly reduced. And meanwhile, the normal vector, curvature and distance information are utilized to further constraint and seek matching point pairs. Then, a random sampling consistency algorithm (RANSAC) is utilized to obtain accurate corresponding point pairs, a singular value decomposition method is adopted to calculate initial registration parameters, and finally, an iterative closest point method (ICP) is utilized to accurately register point clouds.
The specific implementation steps of the application are as follows:
step 1: and acquiring two groups of data of a source point cloud P and a target point cloud Q, requiring that the point cloud is acquired without a visual angle, has a certain overlapping degree, and then calculating a point cloud normal vector. Fig. 4 is a point cloud initial pose diagram.
Step 2: downsampling is performed using voxel grid centroid neighboring points. The method specifically comprises the following steps:
(1) Acquiring a set of keypoints P using voxel filters i 、Q i 。
(2) Traversing the source point cloud P and the target point cloud Q by utilizing KD-Tree, and searching the point set P i 、Q i Nearest neighbor point of each point in the list, forming a new key point set P s 、Q t 。
Step 3: and constructing a 3D descriptor to obtain a list to be matched.
Firstly, multiplying the downsampled source point cloud key points and the target point cloud key points by a local reference coordinate system through the constructed local reference coordinate system, and converting the key points into a new 3D space to form a 3D descriptor. In the new 3D space, each source point cloud key point is processed with a radius of T d And (5) obtaining a list to be matched.
Step 4: and constructing a neighborhood point characteristic histogram descriptor and determining matching point pairs.
Calculating average distance, curvature and change and normal angle sum of the source point cloud key points and the target point cloud key points and the corresponding neighborhood points in the radius neighborhood, encoding the last three dimensions of the six-dimensional feature descriptor, and finally obtaining the neighborhood point feature histogram descriptor. FIG. 3 is a schematic representation of a neighborhood point feature histogram descriptor calculation. In the list to be matched, the accurate corresponding neighbor of the source key point is found in the post three-dimensional point neighborhood point characteristic histogram descriptor through Euclidean measurement, and the accurate matching of the source key point is found from the list to be matched.
Step 5: after the random sampling consistency algorithm is used, mismatching point pairs are removed, and fig. 5 is a point cloud registration result.
According to analysis, the point cloud registration algorithm designed by the algorithm provides a six-dimensional local feature descriptor, and the descriptor has the characteristics of low dimensionality, strong descriptive property, high extraction speed and the like, obviously improves the point cloud registration speed, and has excellent registration precision.
The foregoing detailed description of the application has been presented for purposes of illustration and description, and it should be understood that the application is not limited to the particular embodiments disclosed, but is intended to cover all modifications, equivalents, alternatives, and improvements within the spirit and principles of the application.
Claims (4)
1. A point cloud registration method based on a low-dimensional point cloud local feature descriptor is characterized by comprising the following steps of: the method comprises the following steps:
step 1: acquiring point cloud data and calculating a point cloud data normal vector;
step 2: selecting nearest neighbor point of voxel gravity center as key point K through point cloud downsampling i ；
Step 3: from key point K i Acquiring a Surface neighborhood point set, wherein the Surface neighborhood point set is a neighborhood space Surface with the Surface radius r in point cloud data by using key points ik All points searched;
step 3.1: calculate key point K i Surface radius r neighborhood space Surface ik Inner centroid coordinates mean pt ；
Step 3.2: by Surface treatment ik Subtracting the centroid coordinate mean from all the point coordinates in (a) pt Effectively generates a new surface
Step 3.3: let key point K i Subtracting centroid coordinates mean pt Obtaining
Step 3.4: from a new surfaceA modified covariance matrix COV for creating a local reference coordinate system,
the above formula for calculating the modified covariance matrix is as follows:
wherein q m Represented in a 3D curved surfaceIn (c) q represents->Points d in (a) m ＝||q m -q|| 2 ；
Estimating local reference frame [ RF] 3×3 Will beThe conversion process to a new 3D space is as follows:
wherein [ K ] iRF ] 3×1 New 3D coordinates representing key points [ K ] iRF ] 3×1 The values of each dimension in the low-dimensional point cloud local feature descriptor form a front three-dimensional descriptor of the low-dimensional point cloud local feature descriptor, and the front three-dimensional descriptor is a 3D descriptor;
step 4: calculating the sum of cosine values of included angles between normal vectors of key points and normal vectors of neighborhood space points, the sum of curvature of the key points and curvature values of the neighborhood space points and the sum of distances between the key points and the neighborhood space points to obtain three values, constructing a neighborhood point characteristic histogram descriptor together through the three values, wherein the neighborhood point characteristic histogram descriptor is a rear three-dimensional descriptor, and the 3D descriptor and the neighborhood point characteristic histogram descriptor jointly form a low-dimensional point cloud local characteristic descriptor;
step 5: and matching is carried out through the geometric surface description of the key points, a rotation matrix and a translation matrix are calculated by using a random sampling consistency algorithm, and point cloud registration is completed.
2. The point cloud registration method based on the low-dimensional point cloud local feature descriptors of claim 1, wherein: the step 2 adopts the following steps to extract key points:
step 2.1: establishing KD-Tree for the point cloud, counting the total number N of points in the point cloud, calculating the minimum bounding box of the point cloud and the volume V thereof, and setting L x 、L y 、L z The lengths of the bounding boxes in the x, y and z directions are respectively; the KD-Tree is fully called as K-dimension Tree, is used as a KNN retrieval data structure and is used for storing a K-dimensional space vector point set and carrying out quick retrieval;
step 2.2: setting the side length of the small cube asThe point cloud is divided into n=l×w×h microcubes; wherein: s is a scaling factor, α is a scaling factor for adjusting the side length according to the number of point clouds, l=ceil (L x /L),w＝ceil(L y /L),h＝ceil(L z L), ceil is an upward rounding function;
step 2.3: calculating the gravity center of each non-empty small cube as p 0 (x 0 ,y 0 ,z 0 ) Whereinx i 、y i 、z i X, y, z axis coordinate values for the point cloud of each non-empty small cube;
step 2.4: acquiring two groups of data of a source point cloud P and a target point cloud Q, and acquiring a key point set P by using a voxel filter i 、Q i Traversing the source point cloud P and the target point cloud Q by utilizing KD-Tree, and searching the point set P i 、Q i Nearest neighbor point of each point in the list, forming a new key point set P s 、Q t 。
3. The point cloud registration method based on the low-dimensional point cloud local feature descriptors of claim 1, wherein: and step 4, calculating a neighborhood point characteristic histogram by adopting the following steps:
step 4.1: calculate key point K i The sum of cosine values of included angles between the normal vector and the normal vector of the surface neighborhood point;
step 4.2: calculate key point K i Is the sum of curvature of the surface neighborhood points and curvature values;
step 4.3: calculate key point K i With Surface ik The sum of the distances between the points;
step 4.4: combining the three characteristic values in the steps 4.1, 4.2 and 4.3 to form the neighborhood point characteristic histogram descriptor.
4. The point cloud registration method based on the low-dimensional point cloud local feature descriptors as claimed in claim 2, wherein: the step 5 is to match and calculate the rigid transformation matrix by adopting the following steps:
step 5.1: first for source keypoints K in the new 3D space iRFsource Radial search threshold T d Within the most recent target keypoint list K iRFtarget ；
Step 5.2: finding out an accurate corresponding point by using Euclidean distance and utilizing a neighborhood point characteristic histogram descriptor to generate a matching point pair;
step 5.3: for the corresponding characteristic point pairs, a singular value decomposition algorithm is used for obtaining a three-dimensional rigid transformation matrix from a source point cloud to a target point cloud;
step 5.4: according to the estimated three-dimensional rigid transformation matrix, respectively carrying out coordinate transformation on each pair of corresponding feature points, and obtaining the number of inner points under the three-dimensional rigid transformation matrix according to Euclidean distance between the corresponding feature points after the coordinate transformation;
step 5.5: and 5.4, continuously repeating the step to obtain the number of internal points under each pair of corresponding characteristic point estimation three-dimensional rigid transformation matrixes, determining the estimation three-dimensional rigid transformation matrix with the maximum number of the estimated internal points as a final three-dimensional rigid transformation matrix, carrying out coordinate transformation on the source point cloud, and completing point cloud registration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210611042.3A CN114972459B (en) | 2022-05-31 | 2022-05-31 | Point cloud registration method based on low-dimensional point cloud local feature descriptor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210611042.3A CN114972459B (en) | 2022-05-31 | 2022-05-31 | Point cloud registration method based on low-dimensional point cloud local feature descriptor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114972459A CN114972459A (en) | 2022-08-30 |
| CN114972459B true CN114972459B (en) | 2023-09-08 |
Family
ID=82958184
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210611042.3A Active CN114972459B (en) | 2022-05-31 | 2022-05-31 | Point cloud registration method based on low-dimensional point cloud local feature descriptor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114972459B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115830340B (en) * | 2022-11-16 | 2023-11-21 | 白城师范学院 | Point cloud target identification method and device and storage medium |
| CN116046017B (en) * | 2023-01-18 | 2024-05-07 | 上汽通用五菱汽车股份有限公司 | Calibration method and device for measuring path, storage medium and computer equipment |
| CN115953604B (en) * | 2023-03-13 | 2023-05-30 | 泰安市金土地测绘整理有限公司 | Real estate geographic information mapping data acquisition method |
| CN116434221B (en) * | 2023-06-13 | 2023-09-01 | 中科航迈数控软件(深圳)有限公司 | Workpiece shape recognition method, device, terminal equipment and computer storage medium |
| CN116817771B (en) * | 2023-08-28 | 2023-11-17 | 南京航空航天大学 | Aerospace part coating thickness measurement method based on cylindrical voxel characteristics |
| CN117237428B (en) * | 2023-11-15 | 2024-02-20 | 南京航空航天大学 | Data registration method, device and medium for three-dimensional point cloud |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109215129A (en) * | 2017-07-05 | 2019-01-15 | 中国科学院沈阳自动化研究所 | A kind of method for describing local characteristic based on three-dimensional point cloud |
| CN110335297A (en) * | 2019-06-21 | 2019-10-15 | 华中科技大学 | A kind of point cloud registration method based on feature extraction |
| CN110490912A (en) * | 2019-07-17 | 2019-11-22 | 哈尔滨工程大学 | 3D-RGB point cloud registration method based on local gray level sequence model descriptor |
| CN111080684A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨工程大学 | Point cloud registration method for point neighborhood scale difference description |
| CN112651408A (en) * | 2021-01-07 | 2021-04-13 | 华中科技大学 | Point-to-point transformation characteristic-based three-dimensional local surface description method and system |
| CN114332172A (en) * | 2021-10-14 | 2022-04-12 | 中北大学 | Improved laser point cloud registration method based on covariance matrix |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11080870B2 (en) * | 2019-06-19 | 2021-08-03 | Faro Technologies, Inc. | Method and apparatus for registering three-dimensional point clouds |
-
2022
- 2022-05-31 CN CN202210611042.3A patent/CN114972459B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109215129A (en) * | 2017-07-05 | 2019-01-15 | 中国科学院沈阳自动化研究所 | A kind of method for describing local characteristic based on three-dimensional point cloud |
| CN110335297A (en) * | 2019-06-21 | 2019-10-15 | 华中科技大学 | A kind of point cloud registration method based on feature extraction |
| CN110490912A (en) * | 2019-07-17 | 2019-11-22 | 哈尔滨工程大学 | 3D-RGB point cloud registration method based on local gray level sequence model descriptor |
| CN111080684A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨工程大学 | Point cloud registration method for point neighborhood scale difference description |
| CN112651408A (en) * | 2021-01-07 | 2021-04-13 | 华中科技大学 | Point-to-point transformation characteristic-based three-dimensional local surface description method and system |
| CN114332172A (en) * | 2021-10-14 | 2022-04-12 | 中北大学 | Improved laser point cloud registration method based on covariance matrix |
Non-Patent Citations (1)
| Title |
|---|
| "基于体素下采样和关键点提取的点云自动配准";张彬;《激光与光电子学进展》;第57卷(第4期);第1-9页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114972459A (en) | 2022-08-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114972459B (en) | Point cloud registration method based on low-dimensional point cloud local feature descriptor | |
| CN109887015B (en) | Point cloud automatic registration method based on local curved surface feature histogram | |
| JP6216508B2 (en) | Method for recognition and pose determination of 3D objects in 3D scenes | |
| CN102236794B (en) | Recognition and pose determination of 3D objects in 3D scenes | |
| CN107833181B (en) | Three-dimensional panoramic image generation method based on zoom stereo vision | |
| CN111080684B (en) | Point cloud registration method for point neighborhood scale difference description | |
| CN108376408B (en) | Three-dimensional point cloud data rapid weighting registration method based on curvature features | |
| CN111553409B (en) | Point cloud identification method based on voxel shape descriptor | |
| CN109493372B (en) | Rapid global optimization registration method for product point cloud data with large data volume and few characteristics | |
| CN107818598B (en) | Three-dimensional point cloud map fusion method based on visual correction | |
| CN114677418B (en) | Registration method based on point cloud feature point extraction | |
| CN111145232A (en) | Three-dimensional point cloud automatic registration method based on characteristic information change degree | |
| CN109949350B (en) | Multi-temporal point cloud automatic registration method based on morphological invariant features | |
| CN110930456A (en) | Three-dimensional identification and positioning method of sheet metal part based on PCL point cloud library | |
| CN109523582B (en) | Point cloud coarse registration method considering normal vector and multi-scale sparse features | |
| CN108830888B (en) | Coarse matching method based on improved multi-scale covariance matrix characteristic descriptor | |
| CN110111375B (en) | Image matching gross error elimination method and device under Delaunay triangulation network constraint | |
| CN113628263A (en) | Point cloud registration method based on local curvature and neighbor characteristics thereof | |
| CN115861397A (en) | Point cloud registration method based on improved FPFH-ICP | |
| CN117132630A (en) | Point cloud registration method based on second-order spatial compatibility measurement | |
| CN114648445B (en) | Multi-view high-resolution point cloud splicing method based on feature point extraction and fine registration optimization | |
| CN114463396B (en) | Point cloud registration method utilizing plane shape and topological graph voting | |
| CN115082716A (en) | Multi-source point cloud rough matching algorithm for road fine reconstruction | |
| CN117745780A (en) | Outdoor large scene 3D point cloud registration method based on isolated cluster removal | |
| CN113256693A (en) | Multi-view registration method based on K-means and normal distribution transformation |
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 |