CN116416287A - Point cloud registration method based on feature description of key points and neighborhood points, electronic equipment and storage medium - Google Patents
Point cloud registration method based on feature description of key points and neighborhood points, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN116416287A CN116416287A CN202211487827.0A CN202211487827A CN116416287A CN 116416287 A CN116416287 A CN 116416287A CN 202211487827 A CN202211487827 A CN 202211487827A CN 116416287 A CN116416287 A CN 116416287A
- Authority
- CN
- China
- Prior art keywords
- point
- neighborhood
- initial
- point cloud
- points
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000012216 screening Methods 0.000 claims abstract description 18
- 239000011159 matrix material Substances 0.000 claims abstract description 12
- 230000009466 transformation Effects 0.000 claims abstract description 12
- 230000000007 visual effect Effects 0.000 claims abstract description 5
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 239000013598 vector Substances 0.000 claims description 84
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000005070 sampling Methods 0.000 claims description 6
- 238000005452 bending Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 238000012847 principal component analysis method Methods 0.000 claims description 3
- 229910000831 Steel Inorganic materials 0.000 claims description 2
- 239000000463 material Substances 0.000 claims description 2
- 239000010959 steel Substances 0.000 claims description 2
- 238000000605 extraction Methods 0.000 description 6
- 230000008030 elimination Effects 0.000 description 5
- 238000003379 elimination reaction Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a point cloud registration method based on feature description of key points and neighborhood points, electronic equipment and a storage medium, which comprise the following steps: step 1, preprocessing point cloud data, which comprises the following steps: step 2, initial registration, comprising: extracting key points, calculating feature descriptors, searching initial corresponding relations, screening corresponding relations, solving an initial transformation matrix, calculating a rigid body transformation matrix according to the correct corresponding relations, and rotationally translating point cloud data of different visual angles to the same coordinate system by utilizing the obtained initial rigid body transformation matrix to finish point cloud initial registration; and 3, accurately registering, namely using an initial registration result as an initial value, and performing iterative operation by using an ICP algorithm until the number of iterations or an objective function meets the requirements, so as to finish the cloud registration of the points of different view angles. The method can realize the point cloud data registration with high efficiency and high precision of point clouds with different view angles, and effectively improve the robustness of the point cloud registration under the conditions of data missing and noise.
Description
Technical Field
The invention relates to registration of three-dimensional point cloud data, in particular to a point cloud registration method based on feature description of key points and neighborhood points, electronic equipment and a storage medium.
Background
With the rise of three-dimensional (3D) imaging technology and the continuous development of three-dimensional point cloud data, three-dimensional reconstruction technology is widely applied in many neighborhoods, including robot navigation, geographic mapping, reverse engineering and the like. In the process of scanning and measuring, the object is required to be scanned for multiple times at different angles under the influence of object shielding, environmental factors or errors of measuring tools, and then a plurality of groups of scanning data are registered and spliced together by using a registration algorithm to form a complete point cloud expression of the object. The point cloud registration is to match and overlap the point clouds obtained from different angles to form a complete point cloud. As the most commonly used registration algorithm at present, the iterative closest point algorithm (Iteration Closet Point, ICP) has simple thought and high precision, but the running speed and convergence of the algorithm are greatly dependent on the initial pose of the point cloud, and the objective function is easy to fall into the situation of local optimum. Thus, a common registration scheme is a "coarse-to-fine" strategy, i.e. a good initial position is obtained by coarse registration, and then a fast, accurate registration is achieved using ICP algorithm.
Coarse registration using feature descriptors, which can be defined as a mapping from 3D object space to some high-dimensional vector space, is a currently common method. Because of the irregular nature of three-dimensional point clouds, it is a challenge to design a feature descriptor with high overall performance. In practical cases, the obtained point cloud information is often incomplete, and the global feature descriptors are very sensitive to the information, so that research analysis is performed on the local feature descriptors and the registration algorithm thereof.
Disclosure of Invention
Aiming at the problem that the common registration algorithm is not ideal in multi-view point cloud registration effect, the invention provides a point cloud registration method, electronic equipment and storage medium based on key point and neighborhood point pair feature description, which can realize high-efficiency and high-precision point cloud data registration of point clouds of different view angles and effectively improve the robustness of the point cloud registration under the conditions of data missing and noise.
The invention discloses a point cloud registration method based on feature description of key points and neighborhood points, which comprises the following steps:
step 1, preprocessing point cloud data, which comprises the following steps:
step 1.1, simplifying point cloud data: simplifying point clouds of different visual angles by using a voxel downsampling algorithm;
step 1.2, calculating the algorithm vector and curvature: calculating the normal vector and curvature of each point in the simplified point cloud data by using a principal component analysis method;
step 1.3, redirecting normal vectors;
step 2, initial registration, comprising:
step 2.1, extracting key points: calculating a neighborhood curvature mean value of each point in the point cloud data, and extracting a key point set in the down-sampling point cloud data by utilizing the neighborhood mean value in combination with an ISS algorithm;
step 2.2, feature descriptor calculation: calculating a neighborhood point pair feature vector of the key point set, and calculating a feature descriptor according to the feature vector;
step 2.3, searching initial corresponding relation: calculating the similarity of the neighborhood points to the feature descriptors, and searching the initial corresponding relation bidirectionally according to the similarity;
step 2.4, screening the corresponding relation; firstly, carrying out primary screening on the corresponding relation by using a normal vector included angle mean value, and then carrying out secondary rejection on the corresponding relation by using a RANSAC algorithm to obtain a correct corresponding relation;
step 2.5, solving an initial transformation matrix: calculating a rigid body transformation matrix according to the correct corresponding relation, and rotationally translating the point cloud data of different visual angles to the same coordinate system by using the obtained initial rigid body transformation matrix to finish the initial registration of the point cloud;
and 3, accurately registering, namely using an initial registration result as an initial value, and performing iterative operation by using an ICP algorithm until the number of iterations or an objective function meets the requirements, so as to finish the cloud registration of the points of different view angles.
Further, step 2.1 specifically includes: calculating any point p i Is the neighborhood curvature mean value of (2)
The calculation formula is as follows:
wherein k is p i J=1, 2, …, k, i=1, 2, …, N is the number of point clouds, c ij Curvature for the j-th neighboring point in the i-th point cloud;
will global curvature mean
As a threshold value, screening initial key points;
if it is
Point p i Is an initial key point; if->
Point p i Is not an initial key point;
after the initial key points are extracted, ISS key points in the initial key points are extracted by using an ISS algorithm, and a key point set in the down-sampling point cloud data is obtained.
Further, step 2.2 specifically includes: calculating neighborhood point pair feature vectors of the key points for each key point in the key point set obtained in the step 2.1; for any point p in the point cloud, given the neighborhood radius r, the neighborhood point q in the space sphere can be determined i (i=1, 2, … m); using the same neighborhood radius r to neighbor the point q i (i=1, 2, … m) as a center, m space spheres can be determined; let q i Is q ij (j=1, 2, … n), the neighborhood point-to-feature vector calculation formula of the point p is:
NPFV(p,q i )＝(F 1 ,F 2 ,F 3 ,F 4 ,F 5 ,F 6 ,F 7 ,F 8 )，
wherein q is i Is any neighborhood point of the point p, d i And d ij Are all directed toDifference of d i ＝p-q i ，d ij ＝q i -q ij ；
n is the normal vector of p, n i Is q i Normal vector, n ij Is q ij Normal vector of (2); c p Is the curvature of the p-ray tube,
is q i Is used for the bending of the steel sheet,
is q ij Is a curvature of (2);
F 1 normal vectors n and q for p i Normal vector n of (2) i Included angle between F 2 Normal vectors n and p and q for p i Vector d of constitution i Included angle between F 3 Is q i Normal vector n of (2) i And p and q i D of constitution i Included angle between F 4 P and q i Is used for the bending of the sheet material,
NPFV(p,q i ) P and q i Is a neighborhood point pair feature vector;
after the point-to-point feature vector is obtained, a feature descriptor NPFC (p) is calculated, and the calculation formula is as follows:
where m is the number of p neighborhood points,
the feature vector average value is the neighborhood point pair of the point p.
Further, a calculation formula of similarity of neighbor points of any two points p and q to the feature descriptor is sim= ||NPFC (p) -NPFC (q) |, wherein NPFC (p) is the feature descriptor of the point p, and NPFC (q) is the feature descriptor of the point q.
Further, in step 2.4, the correspondence screening specifically includes: calculating the normal vector included angle of each initial corresponding point pair and the average value of the normal vector included angles of all corresponding point pairs, if the normal vector included angles of the point pairs are smaller than the average value of the normal vector included angles, determining that the corresponding relationship is correct, otherwise, determining that the corresponding relationship is wrong, and thus realizing the primary screening of the corresponding relationship; and then, carrying out secondary corresponding relation screening by using a RANSAC algorithm to obtain a correct corresponding relation.
An electronic device, comprising: a processor and a memory storing a program comprising instructions that when executed by the processor cause the processor to perform the point cloud registration method of the present invention based on a keypoint and neighborhood point-to-feature description.
A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform a point cloud registration method of the present invention based on a keypoint and neighborhood point-to-feature description.
Compared with the prior art, the invention has the following beneficial effects:
1. the key point extraction method adopts a strategy of two-stage extraction, firstly extracts initial key points according to the neighborhood curvature mean value, and then uses an ISS algorithm to carry out secondary key point extraction.
2. The feature descriptor is mainly composed of the space vector included angle and curvature change of the adjacent points, and meanwhile comprises the vector included angle and curvature change mean value of the adjacent point space, and has stronger feature description capability and robustness against noise compared with other common descriptors. The feature descriptors provided by the invention are directly calculated, and the corresponding relation is searched according to the similarity of the feature descriptors, so that the efficiency and the accuracy of feature matching in the point cloud registration process can be effectively improved.
3. Aiming at the initial corresponding relation extracted according to the similarity of the feature descriptors, the invention adopts a strategy of two-stage mismatching elimination, firstly eliminates obvious miscorresponding relation by using a normal vector included angle mean value, and then carries out secondary mismatching elimination by using a RANSAC algorithm. The strategy firstly uses normal vector included angle mean to reject obvious error point pair relation, reduces the iteration times of the RANSAC algorithm, and improves the calculation efficiency of the error matching rejection stage.
Drawings
Fig. 1 is a flowchart of a multi-view point cloud registration method based on a description of key point and neighborhood point pair features according to an exemplary embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, reference will be made to the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention are clearly and completely described, and it is obvious that the described embodiments are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, the point cloud registration method based on the feature description of key points and neighborhood points includes the following steps:
and step 1, preprocessing point cloud data.
Step 1.1, simplifying point cloud data: and setting the side length L of the voxels, and simplifying the point clouds of different view angles by using a voxel downsampling algorithm.
Step 1.2, calculating the algorithm vector and curvature: calculating each point p in the simplified point cloud data by using a principal component analysis method i Normal vector n of (2) i And curvature c i 。
Step 1.3, normal vector redirection, specifically for any point p i The normal vector orientation formula is:
wherein n is i For point p i Normal vector of center (d) pi Is p i Center of mass of neighborhood point cloud pi -p i Is the vector difference.
And 2, initial registration.
And 2, step 2.1, extracting key points: and calculating a neighborhood curvature mean value of each point in the point cloud data, and extracting a key point set in the down-sampling point cloud data by utilizing the neighborhood mean value and an ISS algorithm. The method specifically comprises the following steps: calculating any point p i Is the neighborhood curvature mean value of (2)
The calculation formula is as follows: />
Wherein k is p i J=1, 2, …, k, i=1, 2, …, N is the number of point clouds, c ij Is the curvature of the jth neighboring point in the ith point cloud.
Will global curvature mean
As a threshold value, screening initial key points; if it is
Point p i Is an initial key point; if->
Point p i Is not an initial key point.
After the initial key points are extracted, ISS key points in the initial key points are extracted by using an ISS algorithm, and a key point set in the down-sampling point cloud data is obtained.
The key point extraction method adopts a strategy of two-stage extraction, firstly extracts initial key points according to the neighborhood curvature mean value, and then uses an ISS algorithm to carry out secondary key point extraction.
Step 2.2, feature descriptor calculation: and calculating a neighborhood point pair feature vector of the key point set, and calculating a feature descriptor according to the feature vector. The method comprises the following steps: calculating neighborhood point pair feature vectors of the key points for each key point in the key point set obtained in the step 2.1; for the followingAny point p in the point cloud can determine a neighborhood point q in the space sphere by giving a neighborhood radius r i (i=1, 2, … m); using the same neighborhood radius r to neighbor the point q i (i=1, 2, … m) as a center, m space spheres can be determined; let q i Is q ij (j=1, 2, … n), the neighborhood point-to-feature vector calculation formula of the point p is:
wherein q is i Is any neighborhood point of the point p, d i And d ij Are vector differences, d i ＝p-q i ，d ij ＝q i -q ij The method comprises the steps of carrying out a first treatment on the surface of the n is the normal vector of p, n i Is q i Normal vector, n ij Is q ij Normal vector of (2); c p Is the curvature of the p-ray tube,
is q i Is (are) curvature>
Is q ij Is a curvature of (2); f (F) 1 Normal vectors n and q for p i Normal vector n of (2) i Included angle between F 2 Normal vectors n and p and q for p i Vector d of constitution i Included angle between F 3 Is q i Normal vector n of (2) i And p and q i D of constitution i Included angle between F 4 P and q i Is the difference between the curvatures of NPFV (p, q i ) P and q i Is a neighborhood point pair feature vector.
After the point-to-point feature vector is obtained, a feature descriptor NPFC (p) is calculated, and the calculation formula is as follows:
where m is the number of p neighborhood points,
the feature vector average value is the neighborhood point pair of the point p.
The feature descriptor is mainly composed of the space vector included angle and curvature change of the adjacent points, and meanwhile comprises the vector included angle and curvature change mean value of the adjacent point space, and has stronger feature description capability and robustness against noise compared with other common descriptors. The feature descriptors provided by the invention are directly calculated, and the corresponding relation is searched according to the similarity of the feature descriptors, so that the efficiency and the accuracy of feature matching in the point cloud registration process can be effectively improved.
Step 2.3, searching initial corresponding relation: and calculating the similarity sim of the neighborhood points to the feature descriptors, wherein a calculation formula of the similarity sim of the neighborhood points of any two points p and q to the feature descriptors is sim= ||NPFC (p) -NPFC (q) |, wherein NPFC (p) is the feature descriptor of the point p, and NPFC (q) is the feature descriptor of the point q. And searching the initial corresponding relation bidirectionally according to the similarity.
Step 2.4, screening the corresponding relation; firstly, calculating a normal vector included angle of each initial corresponding point pair and a normal vector included angle average value of all corresponding point pairs, if the normal vector included angle of each point pair is smaller than the normal vector included angle average value, determining that the corresponding relation is correct, otherwise, determining that the corresponding relation is wrong, and accordingly, realizing primary screening of the corresponding relation; and then, carrying out secondary corresponding relation screening by using a RANSAC algorithm to obtain a correct corresponding relation. Namely, the normal vector included angle is as follows for relation screening conditions: if it is
Then the correct correspondence is considered; if->
Then it is considered an erroneous correspondence; wherein n is the number of initial correspondences, i=1, 2, …, n, θ i Is the included angle of the corresponding point to the normal vector. Then using RANSAC algorithm to make the correspondentAnd (5) carrying out secondary elimination on the relationship to obtain a correct corresponding relationship.
Aiming at the initial corresponding relation extracted according to the similarity of the feature descriptors, the invention adopts a strategy of two-stage mismatching elimination, firstly eliminates obvious miscorresponding relation by using a normal vector included angle mean value, and then carries out secondary mismatching elimination by using a RANSAC algorithm. The strategy firstly uses normal vector included angle mean to reject obvious error point pair relation, reduces the iteration times of the RANSAC algorithm, and improves the calculation efficiency of the error matching rejection stage.
Step 2.5, solving an initial transformation matrix: and calculating a rigid body transformation matrix according to the correct corresponding relation, and rotationally translating the point cloud data of different view angles to the same coordinate system by using the obtained initial rigid body transformation matrix to finish the initial registration of the point cloud.
And 3, accurately registering, namely using an initial registration result as an initial value, and performing iterative operation by using an ICP algorithm until the number of iterations or an objective function meets the requirements, so as to finish the cloud registration of the points of different view angles.
An electronic device, comprising: a processor and a memory storing a program comprising instructions that when executed by the processor cause the processor to perform the point cloud registration method of the present invention based on a keypoint and neighborhood point-to-feature description.
A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform a point cloud registration method of the present invention based on a keypoint and neighborhood point-to-feature description.
Claims (7)
1. The point cloud registration method based on the feature description of the key points and the neighborhood points is characterized by comprising the following steps:
step 1, preprocessing point cloud data, which comprises the following steps:
step 1.1, simplifying point cloud data: simplifying point clouds of different visual angles by using a voxel downsampling algorithm;
step 1.2, calculating the algorithm vector and curvature: calculating the normal vector and curvature of each point in the simplified point cloud data by using a principal component analysis method;
step 1.3, redirecting normal vectors;
step 2, initial registration, comprising:
step 2.1, extracting key points: calculating a neighborhood curvature mean value of each point in the point cloud data, and extracting a key point set in the down-sampling point cloud data by utilizing the neighborhood mean value in combination with an ISS algorithm;
step 2.2, feature descriptor calculation: calculating a neighborhood point pair feature vector of the key point set, and calculating a feature descriptor according to the feature vector;
step 2.3, searching initial corresponding relation: calculating the similarity of the neighborhood points to the feature descriptors, and searching the initial corresponding relation bidirectionally according to the similarity;
step 2.4, screening the corresponding relation; firstly, carrying out primary screening on the corresponding relation by using a normal vector included angle mean value, and then carrying out secondary rejection on the corresponding relation by using a RANSAC algorithm to obtain a correct corresponding relation;
step 2.5, solving an initial transformation matrix: calculating an initial rigid body transformation matrix according to the correct corresponding relation, and rotationally translating the point cloud data of different visual angles to the same coordinate system by using the obtained initial rigid body transformation matrix to finish the initial registration of the point cloud;
and 3, accurately registering, namely using an initial registration result as an initial value, and performing iterative operation by using an ICP algorithm until the number of iterations or an objective function meets the requirements, so as to finish the cloud registration of the points of different view angles.
2. The point cloud registration method based on the feature description of the key point and the neighborhood point according to claim 1, wherein step 2.1 specifically comprises: calculating any point p i Is the neighborhood curvature mean value of (2)
The calculation formula is as follows: />
Wherein k is p i J=1, 2, …, k, i=1, 2, …, N is the number of adjacent points of (a)Number of point clouds, c ij Curvature for the j-th neighboring point in the i-th point cloud;
will global curvature mean
As a threshold value, screening initial key points;
if it is
Point p i Is an initial key point; if->
Point p i Is not an initial key point;
after the initial key points are extracted, ISS key points in the initial key points are extracted by using an ISS algorithm, and a key point set in the down-sampling point cloud data is obtained.
3. The point cloud registration method based on the feature description of the key point and the neighborhood point according to claim 1 or 2, wherein step 2.2 specifically comprises: calculating neighborhood point pair feature vectors of the key points for each key point in the key point set obtained in the step 2.1; for any point p in the point cloud, given the neighborhood radius r, the neighborhood point q in the space sphere can be determined i (i=1, 2, … m); using the same neighborhood radius r to neighbor the point q i (i=1, 2, … m) as a center, m space spheres can be determined; let q i Is q ij (j=1, 2, … n), the neighborhood point-to-feature vector calculation formula of the point p is:
NPFV(p，q i )＝(F 1 ，F 2 ，F 3 ，F 4 ，F 5 ，F 6 ，F 7 ，F 8 )，
wherein q is i Is any neighborhood point of the point p, d i And d ij Are vector differences, d i ＝p-q i ，d ij ＝q i -q ij ；
n is the normal vector of p, n i Is q i Normal vector, n ij Is q ij Normal vector of (2); c p Curvature of p, c qi Is q i Is used for the bending of the steel sheet,
is q ij Is a curvature of (2);
F 1 normal vectors n and q for p i Normal vector n of (2) i Included angle between F 2 Normal vectors n and p and q for p i Vector d of constitution i Included angle between F 3 Is q i Normal vector n of (2) i And p and q i D of constitution i Included angle between F 4 P and q i Is used for the bending of the sheet material,
NPFV(p，q i ) P and q i Is a neighborhood point pair feature vector;
after the point-to-point feature vector is obtained, a feature descriptor NPFC (p) is calculated, and the calculation formula is as follows:
where m is the number of p neighborhood points,
the feature vector average value is the neighborhood point pair of the point p.
4. A point cloud registration method based on a feature description of key points and neighborhood points according to claim 3, wherein: the calculation formula of the similarity of the neighborhood points of any two points p and q to the feature descriptor is sim= ||NPFC (p) -NPFC (q) |, wherein NPFC (p) is the feature descriptor of the point p, and NPFC (q) is the feature descriptor of the point q.
5. The point cloud registration method based on the feature description of key points and neighborhood points according to claim 1 or 2, wherein: in step 2.4, the correspondence screening specifically includes: calculating the normal vector included angle of each initial corresponding point pair and the average value of the normal vector included angles of all corresponding point pairs, if the normal vector included angles of the point pairs are smaller than the average value of the normal vector included angles, determining that the corresponding relationship is correct, otherwise, determining that the corresponding relationship is wrong, and thus realizing the primary screening of the corresponding relationship; and then, carrying out secondary corresponding relation screening by using a RANSAC algorithm to obtain a correct corresponding relation.
6. An electronic device, comprising: a processor and a memory storing a program, characterized in that: the program comprises instructions which, when executed by the processor, cause the processor to perform a point cloud registration method based on a keypoint and neighborhood point-to-feature description as claimed in any of claims 1 to 5.
7. A non-transitory computer readable storage medium storing computer instructions, characterized by: the computer instructions are configured to cause the computer to perform the point cloud registration method according to any of claims 1 to 5 based on a keypoint and neighborhood point-to-feature description.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211487827.0A CN116416287A (en) | 2022-11-25 | 2022-11-25 | Point cloud registration method based on feature description of key points and neighborhood points, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211487827.0A CN116416287A (en) | 2022-11-25 | 2022-11-25 | Point cloud registration method based on feature description of key points and neighborhood points, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116416287A true CN116416287A (en) | 2023-07-11 |
Family
ID=87057064
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211487827.0A Pending CN116416287A (en) | 2022-11-25 | 2022-11-25 | Point cloud registration method based on feature description of key points and neighborhood points, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116416287A (en) |
-
2022
- 2022-11-25 CN CN202211487827.0A patent/CN116416287A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lei et al. | Fast descriptors and correspondence propagation for robust global point cloud registration | |
| Yao et al. | Point cloud registration algorithm based on curvature feature similarity | |
| CN108376408B (en) | Three-dimensional point cloud data rapid weighting registration method based on curvature features | |
| CN101311963B (en) | Round mark point center picture projection point position acquiring method for positioning video camera | |
| CN108830888B (en) | Coarse matching method based on improved multi-scale covariance matrix characteristic descriptor | |
| CN103136525B (en) | A kind of special-shaped Extended target high-precision locating method utilizing Generalized Hough Transform | |
| CN107958443B (en) | Fingerprint image splicing method based on ridge line characteristics and TPS deformation model | |
| CN111145232A (en) | Three-dimensional point cloud automatic registration method based on characteristic information change degree | |
| CN111986219B (en) | Matching method of three-dimensional point cloud and free-form surface model | |
| CN112070832B (en) | Non-cooperative target pose tracking method and device based on point cloud DCA (distributed computing architecture) features | |
| US20210271920A1 (en) | Method for grasping texture-less metal parts based on bold image matching | |
| CN110009680B (en) | Monocular image position and posture measuring method based on circle feature and different-surface feature points | |
| CN103400136A (en) | Target identification method based on elastic matching | |
| CN111709450A (en) | Point cloud normal vector estimation method and system based on multi-scale feature fusion | |
| CN113706381A (en) | Three-dimensional point cloud data splicing method and device | |
| CN111199558A (en) | Image matching method based on deep learning | |
| CN115861397A (en) | Point cloud registration method based on improved FPFH-ICP | |
| CN114494368A (en) | Low-overlapping-rate point cloud registration method combining dimensionality reduction projection and feature matching | |
| Xu et al. | GLORN: Strong generalization fully convolutional network for low-overlap point cloud registration | |
| CN113313701A (en) | Electric vehicle charging port two-stage visual detection positioning method based on shape prior | |
| CN116309026A (en) | Point cloud registration method and system based on statistical local feature description and matching | |
| CN110472651A (en) | A kind of object matching and localization method based on marginal point local feature value | |
| CN114358166A (en) | Multi-target positioning method based on self-adaptive k-means clustering | |
| Wen et al. | Recognition and inspection of manufactured parts using line moments of their boundaries | |
| CN117372480A (en) | Point cloud registration method utilizing self-adaptive downsampling and curvature enhancement point-to-point characteristics |
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 |