CN115601408A - Point cloud registration method based on particle swarm optimization and topological graph - Google Patents
Point cloud registration method based on particle swarm optimization and topological graph Download PDFInfo
- Publication number
- CN115601408A CN115601408A CN202211242948.9A CN202211242948A CN115601408A CN 115601408 A CN115601408 A CN 115601408A CN 202211242948 A CN202211242948 A CN 202211242948A CN 115601408 A CN115601408 A CN 115601408A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- registered
- initial
- reference point
- rotation matrix
- 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.)
- Granted
Links
- 239000002245 particle Substances 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000005457 optimization Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 58
- 239000011159 matrix material Substances 0.000 claims abstract description 55
- 238000013519 translation Methods 0.000 claims abstract description 36
- 230000009466 transformation Effects 0.000 claims abstract description 23
- 238000004590 computer program Methods 0.000 claims description 11
- 238000009616 inductively coupled plasma Methods 0.000 claims description 11
- 238000006243 chemical reaction Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 5
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000013178 mathematical model Methods 0.000 description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a point cloud registration method based on particle swarm optimization and a topological graph, which comprises the following steps of: acquiring a reference point cloud and a point cloud to be registered; respectively extracting point cloud characteristics of the reference point cloud and the point cloud characteristics of the point cloud to be registered to form an initial reference point cloud characteristic set and an initial point cloud characteristic set to be registered; respectively obtaining an edge maximization model of the reference point cloud and the point cloud to be registered by using the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered; solving the edge maximization model to obtain an initial rotation matrix and an initial translation vector; optimizing the initial rotation matrix and the initial translation vector to obtain the optimized rotation matrix and translation vector and construct transformation model parameters; and registering the reference point cloud and the point cloud to be registered by using the transformation model parameters. The method utilizes the characteristic that a topological graph relation and a particle swarm algorithm do not need good initialization, has robustness on data containing a large amount of noise, and can realize more stable point cloud registration.
Description
Technical Field
The invention belongs to the technical field of point cloud registration, and particularly relates to a point cloud registration method based on particle swarm optimization and a topological graph.
Background
The point cloud registration technology refers to a process of performing spatial coordinate transformation on two or more point clouds of the same scene acquired at different moments, different visual angles or different sensors to align the two or more point clouds. Point cloud registration is a very key link in three-dimensional data processing, and has been widely applied to the fields of three-dimensional reconstruction, three-dimensional classification, three-dimensional segmentation, automatic driving and the like.
The point cloud registration algorithm is used as a data processing algorithm based on computer vision, and various algorithms with different superior performances have been developed for different scenes through the development of recent decades; iterative Closest Point algorithm (ICP), a milestone algorithm of Point cloud registration algorithm, has become one of the most commonly used Point cloud registration algorithms. Most of the traditional ICP variants are based on geometric features, and the objective function is continuously subjected to iterative optimization so as to minimize the error of the objective function; but ICP often requires a good initial correspondence between pairs of point clouds, otherwise it is easy to get into local optima and the goal of accurate registration is not achieved. In the process of acquiring three-dimensional data by a three-dimensional scanner, due to the influence of factors such as noise, illumination, scanner precision and human interference, a large amount of outliers and noise often exist in the acquired data; although some edge outliers are removed through the processing of operations such as point cloud preprocessing and the like, a large number of outliers still exist in the point cloud; therefore, when registering point cloud data with a large amount of noise and outliers, it is necessary to perform coarse registration on the point cloud data with a large amount of noise and outliers in order to find a good initial correspondence for ICP.
In recent years, a point cloud registration algorithm based on graph optimization is just a research hotspot, the main idea of point cloud registration based on graph matching is to use a non-parametric model to process the point cloud registration problem, and the point cloud registration by utilizing graph matching has the main advantage that topological structures in point cloud data can be fully utilized, so the point cloud registration method based on graph optimization is often robust to data with noise and outliers; since graphs are composed of edges and vertices, graph matching methods aim at finding point correspondences between two graphs by considering the relationships of vertices and edges, and such correspondence search problems in graph matching methods can be considered as optimization problems; researchers can find more accurate corresponding relations by finding better pattern matching optimization strategies. However, most of the existing point cloud registration algorithms based on graph optimization can only realize rough registration and cannot realize precise registration.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a point cloud registration method based on particle swarm optimization and a topological graph. The technical problem to be solved by the invention is realized by the following technical scheme:
the invention provides a point cloud registration method based on particle swarm optimization and a topological graph, which comprises the following steps:
s1: acquiring a reference point cloud and a point cloud to be registered;
s2: respectively extracting point cloud features of the reference point cloud and the point cloud to be registered to form an initial reference point cloud feature set and an initial point cloud feature set to be registered;
s3: respectively obtaining an edge maximization model of the reference point cloud and the point cloud to be registered by using the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered;
s4: solving the edge maximization model by using a particle swarm algorithm as search constraint to obtain an initial rotation matrix and an initial translation vector;
s5: optimizing the initial rotation matrix R and the initial translation vector t by using an ICP (inductively coupled plasma) algorithm to obtain an optimized rotation matrix and an optimized translation vector and construct transformation model parameters;
s6: and registering the reference point cloud and the point cloud to be registered by using the transformation model parameters.
In one embodiment of the present invention, the S2 includes:
s2.1: respectively carrying out filtering operation on the reference point cloud and the point cloud to be registered, and deleting obvious outlier points in the point cloud;
s2.2: respectively performing down-sampling on the filtered reference point cloud and the point cloud to be registered so as to reduce the density of the point cloud;
s2.3: respectively extracting key points of the downsampled reference point cloud and the downsampled point cloud to be registered to obtain a reference key point cloud and a key point cloud to be registered;
s2.4: respectively constructing feature descriptors of the reference key point cloud and the key point cloud to be registered to obtain a reference point cloud feature vector and a point cloud feature vector to be registered;
s2.5: and forming a reference point cloud initial feature set by using the reference point cloud feature vector, and forming a point cloud initial feature set to be registered by using the point cloud feature vector to be registered.
In one embodiment of the present invention, the S3 includes:
s3.1: constructing a feature map G of the reference point cloud based on the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered P And a feature map G of the point cloud to be registered Q ；
S3.2: and converting the corresponding relation between the points of the reference point cloud and the point cloud to be registered into the corresponding relation between edges formed by connecting the points, and constructing an edge maximization model of the reference point cloud and the point cloud to be registered.
In one embodiment of the present invention, said S3.2 comprises:
utilizing the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered to obtain the corresponding relation of edges between the reference point cloud and the point cloud to be registered:
where R denotes a rotation matrix, t denotes a translation vector, p i And p j Point feature vectors representing the ith and jth points in the reference point cloud, q i And q is j Indicates ready to matchPoint feature vectors of the ith point and the jth point corresponding to the quasi-point cloud;
and converting the corresponding relation between every two points of the reference point cloud and the point cloud to be registered into the corresponding relation between edges formed by connecting two points, and constructing an edge maximization model of the reference point cloud and the point cloud to be registered.
In one embodiment of the present invention, the S4 includes:
s4.1: initializing a learning factor of a particle swarm algorithm, and setting the maximum iteration times;
s4.2: randomly generating N rotation matrices R n Using said rotation matrix R n Solving the distance between any two edges in the constructed edge maximization model of the reference point cloud and the point cloud to be registered:
wherein N =1,2., N represents the number of particles in the particle swarm algorithm, and C represents the weight
An nth edge in the set of edge features representing the reference point cloud,
representing the nth edge in the edge feature set of the point cloud to be registered;
s4.3: updating the rotation matrix R n Stopping updating and obtaining an initial rotation matrix R until the formula requirement of the step S4.2 is met or the maximum iteration number is reached, wherein the rotation matrix R is updated n The formula of (1) is:
R n ＝R n +(R P -R n )×C 1 ×r 1n +(R g -R n )×C 2 ×r 2n ,n＝1,2,...,N
wherein, C 1 And C 2 Denotes the acceleration constant, r 1i Represents a random number of 0 to 1; r is 2i Represents a random number of 0 to 1;
s4.4: and obtaining an initial translation vector t of the reference point cloud and the point cloud to be registered by utilizing a maximum consensus thought according to the obtained initial rotation matrix R.
In one embodiment of the present invention, the S5 includes:
s5.1: performing initial conversion on the reference point cloud by using the initial rotation matrix R and the initial translation vector t to obtain a converted reference point cloud P';
s5.2: traversing each point in the converted reference point cloud P ', and searching each point P' in the converted reference point cloud P 'in the point cloud to be registered' i The point with the minimum Euclidean distance is combined into a corresponding point set;
s5.3: using SVD to calculate rigid transformation to minimize the residual square sum d of the corresponding point set to obtain an updated rotation matrix R 'and a translation matrix t';
s5.4: transforming the point cloud Q to be registered by using a rotation matrix R ' and a translation matrix t ' to obtain a transformed point cloud Q ';
s5.5: judging whether the sum of the squares of the residual errors of the point sets corresponding to the point clouds Q 'and P' after transformation is smaller than a given threshold value, if so, terminating iterative transformation, and outputting a final rotation matrix R and a final translation vector t; if the point cloud Q 'is larger than the original point cloud Q to be registered, the transformed point cloud Q' is used for replacing the original point cloud Q to be registered, and the steps S5.2-S5.5 are repeated to continue iteration.
Another aspect of the present invention provides a storage medium, in which a computer program is stored, the computer program being configured to execute the steps of the particle swarm optimization and topology map-based point cloud registration method described in any one of the above embodiments.
Yet another aspect of the present invention provides an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor, when invoking the computer program in the memory, implements the steps of the particle swarm optimization and topology map-based point cloud registration method according to any one of the above embodiments.
Compared with the prior art, the invention has the beneficial effects that:
1. the point cloud registration method can convert the corresponding relation between the solved points into the corresponding relation between the solved edges by utilizing the topological structure among point cloud data based on the particle swarm optimization algorithm, establish a mathematical model with the edge corresponding maximization, and solve the edge maximization model by taking the particle swarm optimization as search constraint to obtain an initial conversion relation; and iterative optimization is carried out on the initial conversion relation by utilizing an ICP algorithm to obtain an optimal rotation matrix and translation vector, and the registration precision and efficiency of the point cloud can be enhanced by introducing a particle swarm optimization algorithm.
2. The point cloud registration method provided by the invention utilizes the characteristic that good initialization is not needed by a topological graph relation and a particle swarm algorithm, has robustness on data containing a large amount of noise, and can realize more stable point cloud registration.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Drawings
Fig. 1 is a flowchart of a point cloud registration method based on particle swarm optimization and a topological graph according to an embodiment of the present invention.
Detailed Description
In order to further illustrate the technical means and efficacy adopted by the present invention to achieve the predetermined invention purpose, a point cloud registration method based on particle swarm optimization and a topological graph according to the present invention is described in detail below with reference to the accompanying drawings and the detailed embodiments.
The foregoing and other technical matters, features and effects of the present invention will be apparent from the following detailed description of the embodiments, which is to be read in connection with the accompanying drawings. The technical means and effects of the present invention adopted to achieve the predetermined purpose can be more deeply and specifically understood through the description of the specific embodiments, however, the attached drawings are provided for reference and description only and are not used for limiting the technical scheme of the present invention.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that an article or device that comprises a list of elements does not include only those elements but may include other elements not expressly listed. Without further limitation, an element defined by the phrase "comprising a … …" does not exclude the presence of additional like elements in an article or apparatus that comprises the element.
Referring to fig. 1, fig. 1 is a schematic flow chart of a point cloud registration method based on particle swarm optimization and a topological graph according to an embodiment of the present invention, where the point cloud registration method includes:
s1: and acquiring a reference point cloud and a point cloud to be registered.
In this embodiment, first, two point clouds are obtained, one of the point clouds is used as a reference point cloud, and the other point cloud is used as a point cloud to be registered. The reference point cloud is a target template point cloud, and the point cloud to be registered is a point cloud which needs to be processed to be coincident with the target template point cloud after coordinate change.
S2: and respectively extracting point cloud features of the reference point cloud and the point cloud features of the point cloud to be registered to form an initial reference point cloud feature set and an initial point cloud feature set to be registered.
Specifically, S2 of the present embodiment includes:
s2.1: and respectively carrying out filtering operation on the reference point cloud and the point cloud to be registered, and deleting obvious outlier points in the point cloud data.
S2.2: and respectively carrying out down-sampling operation on the filtered reference point cloud and the point cloud to be registered so as to reduce the density of the point clouds and further reduce the data volume and algorithm complexity of related processing.
In this embodiment, the existing uniform down-sampling may be adopted to perform down-sampling operations on the filtered reference point cloud and the point cloud to be registered, respectively.
S2.3: and respectively carrying out key point extraction operation on the downsampled reference point cloud and the point cloud to be registered, and selecting a part of points with rich information and strong representativeness to represent the whole point cloud so as to obtain the reference key point cloud and the point cloud to be registered.
In this embodiment, an Intrinsic Shape features (ISS) detector is used to extract key points of the reference point cloud and the point cloud to be registered after the downsampling operation. The specific selection criteria are according to the criteria given in the ISS.
In other embodiments, other suitable key point extraction methods may be used to extract key points from the reference point cloud and the point cloud to be registered.
S2.4: and respectively carrying out feature descriptor construction operation on the reference key point cloud and the key point cloud to be registered to obtain a reference point cloud feature vector and a point cloud feature vector to be registered.
Specifically, feature descriptor construction operation is respectively carried out on the reference key point cloud and the key point cloud to be registered, and key points are coded into feature vectors through descriptors, so that a reference point cloud feature vector and a point cloud feature vector to be registered are obtained, and the feature vectors are used for calculating the matching relationship between two groups of point cloud features.
S2.5: and forming a reference point cloud initial feature set by using the reference point cloud feature vector, and forming a point cloud initial feature set to be registered by using the point cloud feature vector to be registered.
S3: and respectively obtaining an edge maximization model of the reference point cloud and the point cloud to be registered by using the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered.
In this embodiment, in order to eliminate the influence of outliers in the point cloud by using the correlation in the point cloud feature set as much as possible for the case where the noise data contains a large number of outliers, the similarity between the original reference point cloud and the corresponding point in the point cloud to be registered is converted into the similarity of the corresponding edges connected by the points.
In the embodiment, an edge vector is constructed by calculating the difference of coordinates of feature vectors of two points in the same point cloud, namely, the difference of coordinates of feature vectors of every two points in the same point cloud is calculated, the edge vector between every two points is constructed, so that a topological structure of the point cloud is obtained, the corresponding relation between the reference point cloud and the cloud center point of the point to be registered is converted into the corresponding relation between the edges by using the topological structure between the point cloud data, an edge maximization mathematical model is established, and then the problem of solving 6 degrees of freedom is decomposed into the problem of solving 2 degrees of freedom and 3 degrees of freedom.
Specifically, step S3 includes:
s3.1: constructing a feature map G of the reference point cloud based on the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered P And a feature map G of the point cloud to be registered Q 。
Constructing two images, namely a feature image G of the reference point cloud based on the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered P (V P ,E P ) And a feature map G of the point cloud to be registered Q (V Q ,E Q ) Wherein, V P Set of point features representing a reference point cloud, E P Set of edge features, V, representing a reference point cloud Q Set of point features representing the point cloud to be registered, E Q Set of edge features representing the point cloud to be registered, G P Topological model (feature map), G, representing a reference point cloud Q A topological model (feature map) representing the point cloud to be registered.
S3.2: and converting the corresponding relation between the points of the reference point cloud and the point cloud to be registered into the corresponding relation between edges formed by connecting the points, and constructing an edge maximization model of the reference point cloud and an edge maximization model of the point cloud to be registered.
Point feature set V due to reference point cloud P Point characteristic set V of point cloud to be registered Q There is a corresponding relationship between them, therefore
Belonging to a set of point identities, wherein,
set of representative point features V P The point (i) of (a) is,
feature set V representing points Q Neutral point
The corresponding ith point.
If two sets of points
And
all are point consistency sets, and if any two points can determine one edge, the point characteristic set V is used P Point i in
And the j point
Joined edge
And the point feature set V Q Point i in
And the j point
Joined edge
There is also a correct matching relationship; the correspondence between points can thus be converted into a correspondence between edges, wherein,
set of representative point features V P At the j-th point in the (c),
set of representative point features V Q Neutral point
The corresponding point of the j-th point,
represents the edge between the ith point and the jth point in the point feature set P,
representing the edge between the ith point and the jth point in the point feature set Q.
where R denotes a rotation matrix, t denotes a translation vector, p i And p j Point feature vectors representing the ith and jth points in the reference point cloud, q i And q is j And representing point feature vectors of the ith point and the jth point corresponding to the point cloud to be registered.
And according to the relational expression, converting the corresponding relation between all points of the reference point cloud and the point cloud to be registered into the corresponding relation between edges formed by connecting two points, and constructing an edge maximization model of the reference point cloud and the point cloud to be registered, so that the step decomposes the problem of solving 6 degrees of freedom into the problem of solving 2 degrees of freedom and 3 degrees of freedom.
S4: and solving the edge maximization model by using a particle swarm algorithm as a search constraint to obtain an initial rotation matrix and an initial translation vector.
Specifically, the S4 includes:
s4.1: and initializing a learning factor of the particle swarm algorithm, and setting the maximum iteration times.
S4.2: randomly generating N rotation matrices R n Using said rotation matrix R n Solving the distance between any two edges in the constructed edge maximization model of the reference point cloud and the point cloud to be registered:
wherein N =1,2., N represents the number of particles in the particle swarm algorithm, C represents the weight, here set to 1,
an nth edge in the set of edge features representing the reference point cloud,
and representing the nth edge in the edge feature set of the point cloud to be registered.
Corresponding rotation matrix R obtained according to the formula n I.e. the best choice for this round, while setting the historically best position R of the particles p And the optimum position R of all particles g 。
S4.3: updating the rotation matrix R according to the following equation (3) n Satisfying the requirement of step S4.2 or up to the maximum number of iterations, and obtaining an initial rotation matrix R, wherein the rotation matrix R is updated n The formula of (1) is:
R n ＝R n +(R P -R n )×C 1 ×r 1n +(R g -R n )×C 2 ×r 2n ,n＝1,2,...,N (3)
wherein, C 1 And C 2 Denotes the acceleration constant, r 1i A random number representing 0 to 1; r is 2i Represents a random number of 0 to 1. Specifically, C 1 And C 2 It can be adjusted empirically to represent the weights that play a role in the motion.
S4.4: and obtaining an initial translation vector t of the reference point cloud and the point cloud to be registered by utilizing a maximum consensus thought according to the obtained initial rotation matrix R, namely obtaining the obtained initial conversion relation (the initial rotation matrix R and the initial translation vector t).
S5: and optimizing the initial rotation matrix R and the initial translation vector t by utilizing an ICP (inductively coupled plasma) algorithm to obtain an optimized rotation matrix and an optimized translation vector.
In this embodiment, the S5 includes:
s5.1: and performing initial conversion on the reference point cloud by using the initial rotation matrix R and the initial translation vector t to obtain a converted reference point cloud P'.
S5.2: traversing each point in the converted reference point cloud P ', and searching each point P' in the converted reference point cloud P 'in the point cloud to be registered' i And the points with the minimum Euclidean distance correspondence relationship form a corresponding point set.
Specifically, each point in the converted reference point cloud P 'is traversed, and a point with the minimum euclidean distance to each point in the reference point cloud P' is found in the point cloud to be registered, so that pairs of points are formed, and a point set consisting of corresponding points is formed.
S5.3: using SVD (Singular Value Decomposition ) to calculate a rigid transformation to minimize the residual squared sum d of the corresponding point set, resulting in an updated rotation matrix R 'and translation matrix t'. The specific process is the prior art and is not described herein again.
S5.4: and transforming the point cloud Q to be registered by using the rotation matrix R ' and the translation matrix t ' to obtain transformed point cloud Q '.
S5.5: judging whether the sum of the squares of the residual errors of the point sets corresponding to the point clouds Q 'and P' after transformation is smaller than a given threshold value, if so, terminating iterative transformation, and outputting a final rotation matrix R and a final translation vector t; if the point cloud Q 'is larger than the original point cloud Q to be registered, the transformed point cloud Q' is used for replacing the original point cloud Q to be registered, and the steps S5.2-S5.5 are repeated to continue iteration.
S5.6: and obtaining transformation model parameters according to the final rotation matrix R and the translation vector t.
In this embodiment, affine transformation parameters can be obtained by performing singular value decomposition on the final rotation matrix R and the final translation vector t, and the affine transformation parameters are transformation model parameters. In other embodiments, the rotation matrix R and the translation vector t may be processed as transformation model parameters by other methods, which are not limited herein.
S6: and registering the reference point cloud and the point cloud to be registered according to the obtained transformation model parameters.
The point cloud registration method based on particle swarm optimization and the topological graph abstracts the point cloud registration problem into an optimization problem, and realizes a point cloud registration scheme from coarse to fine. In the coarse registration stage, in order to eliminate the influence of outliers in a point set by utilizing the correlation among the point sets as much as possible aiming at the situation that noise data contains a large number of outliers, the similarity between original corresponding points is converted into the similarity of corresponding edges; constructing edge vectors by solving the coordinate difference of two point vectors in the same point cloud, converting the corresponding relation between the solved points into the corresponding relation between the solved edges by using the topological structure between the point cloud data, establishing a mathematical model with maximized edge correspondence, and decomposing the problem of solving 6 degrees of freedom into the problem of solving two 3 degrees of freedom; then, in order to improve the efficiency of global search, the embodiment of the invention uses a particle swarm algorithm as search constraint to solve the edge maximization model and obtain a good initial corresponding relation; and in the fine registration stage, optimizing the initial conversion relation by using an iterative closest point algorithm to obtain a final conversion relation.
Compared with the prior art, the method provided by the embodiment of the invention is superior to the prior art, and particularly, the method utilizes the characteristic that a topological graph relation and a particle swarm algorithm do not need good initialization, has robustness on data containing a large amount of noise, and can realize more robust point cloud registration.
Yet another embodiment of the present invention provides a storage medium, in which a computer program is stored, the computer program being used to execute the steps of the point cloud registration method based on particle swarm optimization and topological graph in the above embodiments. Yet another aspect of the present invention provides an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor, when invoking the computer program in the memory, implements the steps of the particle swarm optimization and topology map-based point cloud registration method according to the above embodiments. Specifically, the integrated module implemented in the form of a software functional module may be stored in a computer readable storage medium. The software functional module is stored in a storage medium and includes several instructions to enable an electronic device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.
Claims (8)
1. A point cloud registration method based on particle swarm optimization and topological graph is characterized by comprising the following steps:
s1: acquiring a reference point cloud and a point cloud to be registered;
s2: respectively extracting point cloud features of the reference point cloud and the point cloud to be registered to form an initial reference point cloud feature set and an initial point cloud feature set to be registered;
s3: respectively obtaining an edge maximization model of the reference point cloud and the point cloud to be registered by using the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered;
s4: solving the edge maximization model by using a particle swarm algorithm as search constraint to obtain an initial rotation matrix and an initial translation vector;
s5: optimizing the initial rotation matrix R and the initial translation vector t by using an ICP (inductively coupled plasma) algorithm to obtain an optimized rotation matrix and an optimized translation vector and construct transformation model parameters;
s6: and registering the reference point cloud and the point cloud to be registered by using the transformation model parameters.
2. The particle swarm optimization and topology map-based point cloud registration method according to claim 1, wherein the S2 comprises:
s2.1: respectively carrying out filtering operation on the reference point cloud and the point cloud to be registered, and deleting obvious outlier points in the point cloud;
s2.2: respectively performing down-sampling on the filtered reference point cloud and the point cloud to be registered so as to reduce the density of the point cloud;
s2.3: respectively extracting key points of the downsampled reference point cloud and the downsampled point cloud to be registered to obtain a reference key point cloud and a key point cloud to be registered;
s2.4: respectively constructing feature descriptors of the reference key point cloud and the key point cloud to be registered to obtain a reference point cloud feature vector and a point cloud feature vector to be registered;
s2.5: and forming a reference point cloud initial feature set by using the reference point cloud feature vector, and forming a point cloud initial feature set to be registered by using the point cloud feature vector to be registered.
3. The particle swarm optimization and topology map-based point cloud registration method according to claim 1, wherein the S3 comprises:
s3.1: constructing a feature map G of the reference point cloud based on the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered P And a feature map G of the point cloud to be registered Q ；
S3.2: and converting the corresponding relation between the points of the reference point cloud and the point cloud to be registered into the corresponding relation between edges formed by connecting the points, and constructing an edge maximization model of the reference point cloud and the point cloud to be registered.
4. The particle swarm optimization and topology map based point cloud registration method of claim 3, wherein S3.2 comprises:
utilizing the initial feature set of the reference point cloud and the initial feature set of the point cloud to be registered to obtain the corresponding relation of edges between the reference point cloud and the point cloud to be registered:
where R denotes a rotation matrix, t denotes a translation vector, p i And p j Point feature vectors representing the ith and jth points in the reference point cloud, q i And q is j Point feature vectors representing the ith point and the jth point corresponding to the point cloud to be registered;
and converting the corresponding relation between every two points of the reference point cloud and the point cloud to be registered into the corresponding relation between edges formed by connecting two points, and constructing an edge maximization model of the reference point cloud and the point cloud to be registered.
5. The particle swarm optimization and topology map-based point cloud registration method according to claim 1, wherein the S4 comprises:
s4.1: initializing a learning factor of a particle swarm algorithm, and setting the maximum iteration times;
s4.2: randomly generating N rotation matrices R n Using said rotation matrix R n Solving the distance between any two edges in the constructed reference point cloud and the edge maximization model of the point cloud to be registered:
wherein N =1,2., N represents the number of particles in the particle swarm algorithm, C represents the weight,
an nth edge in the set of edge features representing the reference point cloud,
representing the nth edge in the edge feature set of the point cloud to be registered;
s4.3: updating the rotation matrix R n Stopping updating and obtaining an initial rotation matrix R until the formula requirement of the step S4.2 is met or the maximum iteration number is reached, wherein the rotation matrix R is updated n The formula of (1) is:
R n ＝R n +(R P -R n )×C 1 ×r 1n +(R g -R n )×C 2 ×r 2n ,n＝1,2,...,N
wherein, C 1 And C 2 Denotes the acceleration constant, r 1i Represents a random number of 0 to 1; r is 2i Represents a random number of 0 to 1;
s4.4: and obtaining an initial translation vector t of the reference point cloud and the point cloud to be registered by utilizing a maximum consensus thought according to the obtained initial rotation matrix R.
6. The particle swarm optimization and topology map-based point cloud registration method according to claim 5, wherein the S5 comprises:
s5.1: performing initial conversion on the reference point cloud by using the initial rotation matrix R and the initial translation vector t to obtain a converted reference point cloud P';
s5.2: traversing each point in the converted reference point cloud P ', and searching each point P' in the converted reference point cloud P 'in the point cloud to be registered' i The point with the minimum Euclidean distance is combined into a corresponding point set;
s5.3: using SVD to calculate rigid transformation to minimize the residual square sum d of the corresponding point set to obtain an updated rotation matrix R 'and a translation matrix t';
s5.4: transforming the point cloud Q to be registered by using a rotation matrix R ' and a translation matrix t ' to obtain a transformed point cloud Q ';
s5.5: judging whether the sum of the squares of the residual errors of the point sets corresponding to the point clouds Q 'and P' after transformation is smaller than a given threshold value, if so, terminating iterative transformation, and outputting a final rotation matrix R and a final translation vector t; if so, replacing the original point cloud Q to be registered with the transformed point cloud Q', and repeating the steps S5.2-S5.5 to continue iteration;
s5.6: and obtaining transformation model parameters according to the final rotation matrix R and the translation vector t.
7. A storage medium, characterized in that the storage medium has stored therein a computer program for performing the steps of the particle swarm optimization and topology map based point cloud registration method of any of claims 1 to 6.
8. An electronic device, comprising a memory having a computer program stored therein and a processor implementing the steps of the particle swarm optimization and topology map based point cloud registration method according to any one of claims 1 to 6 when the processor invokes the computer program in the memory.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211242948.9A CN115601408B (en) | 2022-10-11 | 2022-10-11 | Point cloud registration method based on particle swarm optimization and topological graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211242948.9A CN115601408B (en) | 2022-10-11 | 2022-10-11 | Point cloud registration method based on particle swarm optimization and topological graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115601408A true CN115601408A (en) | 2023-01-13 |
| CN115601408B CN115601408B (en) | 2023-12-12 |
Family
ID=84847339
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211242948.9A Active CN115601408B (en) | 2022-10-11 | 2022-10-11 | Point cloud registration method based on particle swarm optimization and topological graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115601408B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116342666A (en) * | 2023-02-10 | 2023-06-27 | 西安电子科技大学 | Three-dimensional point cloud registration method based on multi-form optimization and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105551015A (en) * | 2015-12-02 | 2016-05-04 | 南京邮电大学 | Scattered-point cloud image registering method |
| CN111260704A (en) * | 2020-01-09 | 2020-06-09 | 北京理工大学 | Vascular structure 3D/2D rigid registration method and device based on heuristic tree search |
| CN111860520A (en) * | 2020-07-21 | 2020-10-30 | 南京航空航天大学 | Large airplane point cloud model self-supervision semantic segmentation method based on deep learning |
| US11037346B1 (en) * | 2020-04-29 | 2021-06-15 | Nanjing University Of Aeronautics And Astronautics | Multi-station scanning global point cloud registration method based on graph optimization |
-
2022
- 2022-10-11 CN CN202211242948.9A patent/CN115601408B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105551015A (en) * | 2015-12-02 | 2016-05-04 | 南京邮电大学 | Scattered-point cloud image registering method |
| CN111260704A (en) * | 2020-01-09 | 2020-06-09 | 北京理工大学 | Vascular structure 3D/2D rigid registration method and device based on heuristic tree search |
| US11037346B1 (en) * | 2020-04-29 | 2021-06-15 | Nanjing University Of Aeronautics And Astronautics | Multi-station scanning global point cloud registration method based on graph optimization |
| CN111860520A (en) * | 2020-07-21 | 2020-10-30 | 南京航空航天大学 | Large airplane point cloud model self-supervision semantic segmentation method based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| LI DING 等: "《DeepMapping: Unsupervised Map Estimation From Multiple Point Clouds》", 《2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)》, pages 1 - 10 * |
| 孙晓鹏: "《耳廓点云形状特征匹配的路径跟随算法》", 《软件学报》, vol. 26, no. 05, pages 1251 - 1264 * |
| 杨佳琪 等: "《多视图点云配准算法综述》", 《华中科技大学学报(自然科学版)》, vol. 50, no. 11, pages 16 - 34 * |
| 王瑜 等: "《基于改进粒子群算法的散乱点云数据配准》", 《科技通报》, vol. 32, no. 10, pages 1 - 5 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116342666A (en) * | 2023-02-10 | 2023-06-27 | 西安电子科技大学 | Three-dimensional point cloud registration method based on multi-form optimization and electronic equipment |
| CN116342666B (en) * | 2023-02-10 | 2024-03-19 | 西安电子科技大学 | Three-dimensional point cloud registration method based on multi-form optimization and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115601408B (en) | 2023-12-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112017220B (en) | Point cloud accurate registration method based on robust constraint least square algorithm | |
| US7382897B2 (en) | Multi-image feature matching using multi-scale oriented patches | |
| US9141871B2 (en) | Systems, methods, and software implementing affine-invariant feature detection implementing iterative searching of an affine space | |
| CN103136520B (en) | The form fit of Based PC A-SC algorithm and target identification method | |
| CN113139996B (en) | Point cloud registration method and system based on three-dimensional point cloud geometric feature learning | |
| CN114677418B (en) | Registration method based on point cloud feature point extraction | |
| CN104090972A (en) | Image feature extraction and similarity measurement method used for three-dimensional city model retrieval | |
| Xu et al. | Iterative manifold embedding layer learned by incomplete data for large-scale image retrieval | |
| CN110807828A (en) | Oblique photography three-dimensional reconstruction matching method | |
| Zhang et al. | DDRNet: Fast point cloud registration network for large-scale scenes | |
| CN113515656A (en) | Multi-view target identification and retrieval method and device based on incremental learning | |
| CN115601408B (en) | Point cloud registration method based on particle swarm optimization and topological graph | |
| CN112801945A (en) | Depth Gaussian mixture model skull registration method based on dual attention mechanism feature extraction | |
| CN113902779B (en) | Point cloud registration method based on tensor voting method | |
| Liao et al. | Fuzzypsreg: Strategies of fuzzy cluster-based point set registration | |
| CN114358166A (en) | Multi-target positioning method based on self-adaptive k-means clustering | |
| CN114332168A (en) | Pedestrian tracking method, pedestrian tracking system, electronic device and storage medium | |
| CN114332172A (en) | Improved laser point cloud registration method based on covariance matrix | |
| CN115661218B (en) | Virtual super-point-based laser point cloud registration method and system | |
| CN110070626B (en) | Three-dimensional object retrieval method based on multi-view classification | |
| Yao et al. | Matching wide-baseline stereo images with weak texture using the perspective invariant local feature transformer | |
| CN111160433B (en) | High-speed matching method and system for high-resolution image feature points | |
| Wang et al. | LiDAR-SLAM loop closure detection based on multi-scale point cloud feature transformer | |
| Makadia | Feature tracking for wide-baseline image retrieval | |
| Liu et al. | An improved local descriptor based object recognition in cluttered 3D point clouds |
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 |