CN113706381A - Three-dimensional point cloud data splicing method and device - Google Patents
Three-dimensional point cloud data splicing method and device Download PDFInfo
- Publication number
- CN113706381A CN113706381A CN202110993313.1A CN202110993313A CN113706381A CN 113706381 A CN113706381 A CN 113706381A CN 202110993313 A CN202110993313 A CN 202110993313A CN 113706381 A CN113706381 A CN 113706381A
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- splicing
- clouds
- voxel
- 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 95
- 239000013598 vector Substances 0.000 claims abstract description 48
- 239000011159 matrix material Substances 0.000 claims abstract description 19
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 13
- 230000009466 transformation Effects 0.000 claims abstract description 13
- 238000001914 filtration Methods 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 9
- 230000005484 gravity Effects 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 2
- 238000002474 experimental method Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012847 principal component analysis method Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4038—Image mosaicing, e.g. composing plane images from plane sub-images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/32—Indexing scheme for image data processing or generation, in general involving image mosaicing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Generation (AREA)
Abstract
The method and the device for splicing the three-dimensional point cloud data can solve the problems of high splicing difficulty and low precision of two pieces of point clouds with low overlapping rate, and effectively improve the splicing precision and the time efficiency of the point cloud data with low initial overlapping degree. The method comprises the following steps: (1) the point cloud acquisition system acquires partially overlapped source point cloud and target point cloud, and preprocesses the point cloud; (2) the method comprises the steps of obtaining good initial positions of a source point cloud and a target point cloud through coordinate transformation of scanning space positions twice; (3) calculating curvature characteristics of the point cloud, accelerating the searching process of nearest neighbor by using a Kdtree to find an initial matching point pair, and setting a constraint condition by using a direction vector threshold method to eliminate an error point pair; (4) and obtaining a rotation matrix R and a translational vector t between the two point clouds by using a Singular Value Decomposition (SVD) method, and completing splicing.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a three-dimensional point cloud data splicing method and a three-dimensional point cloud data splicing device.
Background
Point cloud splicing is an important part in the three-dimensional point cloud processing process, and is a very key task in the fields of reverse engineering, virtual reality, cultural relic restoration and the like. In actual measurement, due to the limitation of the visual angles of a camera and a projector, the complete appearance of an object is difficult to obtain through the measurement of one visual angle, so that the measurement is required to be carried out from different visual angles, the point cloud data obtained under each visual angle is spliced into a complete three-dimensional model, and the process is the splicing of the point clouds.
The current common splicing methods include a splicing method depending on external auxiliary information and a splicing method depending on measurement data. The surface marking method is a splicing method relying on external auxiliary information, and coordinates conversion relation between marking points in initial point cloud and marking points in target point cloud is obtained through the marking points pasted on the surface of a measured object, so that point cloud data under different viewing angles are spliced, but the application range of the method is limited. Splicing methods depending on the measurement data themselves include an iterative closest point method (ICP), a principal component analysis method (PCA), and the like. The ICP algorithm searches the closest point between two pieces of point cloud by iteration and finds the conversion relation when the Euclidean distance reaches the minimum. However, there are two preconditions for using the ICP algorithm, one is that the two point clouds should be sufficiently similar, and the two point clouds need to have good initial positions. In addition, the convergence rate of conventional ICP is typically low and the algorithm requires a large amount of computation time. The PCA firstly calculates the characteristic vector of the point cloud through the covariance matrix, then calculates the main direction of the point cloud and two secondary directions perpendicular to the main direction of the point cloud through the characteristic vector, and completes the splicing of the point cloud by utilizing the three directions. This method is computationally efficient, but requires a sufficiently high overlap of adjacent point clouds.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a three-dimensional point cloud data splicing method, which can solve the problems of high splicing difficulty and low precision of two pieces of point clouds with low overlapping rate and effectively improve the splicing precision and time efficiency of the point cloud data with low initial overlapping degree.
The technical scheme of the invention is as follows: the three-dimensional point cloud data splicing method comprises the following steps:
(1) the point cloud acquisition system acquires partially overlapped source point cloud and target point cloud, and preprocesses the point cloud;
(2) the method comprises the steps of obtaining good initial positions of a source point cloud and a target point cloud through coordinate transformation of scanning space positions twice;
(3) calculating curvature characteristics of the point cloud, accelerating the searching process of nearest neighbor by using a Kdtree to find an initial matching point pair, and setting a constraint condition by using a direction vector threshold method to eliminate an error point pair;
(4) and obtaining a rotation matrix R and a translational vector t between the two point clouds by using a Singular Value Decomposition (SVD) method, and completing splicing.
Firstly, acquiring a source point cloud and a target point cloud which are partially overlapped through a point cloud acquisition system, and preprocessing the point clouds; then, the source point cloud and the target point cloud are enabled to obtain good initial positions through the coordinate transformation of scanning space positions twice; acquiring initial matching point pairs of two point clouds by calculating curvature characteristics of the point clouds and using a Kdtree nearest neighbor searching method, and setting constraint conditions by a direction vector threshold method to eliminate wrong point pairs; and finally, obtaining a rotation matrix R and a translational vector t between the two point clouds by utilizing singular value decomposition. Therefore, the problems of high splicing difficulty and low precision of two pieces of point cloud with low overlapping rate can be solved, and the splicing precision and the time efficiency of the point cloud data with low initial overlapping degree are effectively improved.
Still provide a splicing apparatus of three-dimensional point cloud data, it includes:
the data preprocessing module is configured to acquire partially overlapped source point clouds and target point clouds through the point cloud acquisition system and preprocess the point clouds;
an initial position acquisition module configured to obtain good initial positions of the source point cloud and the target point cloud through coordinate transformation of scanning spatial positions twice;
the data eliminating module is configured to calculate the curvature characteristics of the point cloud, accelerate the searching process of nearest neighbor by using Kdtree to search for an initial matching point pair, and then set a constraint condition by using a direction vector threshold method to eliminate an error point pair;
and the splicing module is configured to obtain a rotation matrix R and a translational vector t between the two point clouds by utilizing a Singular Value Decomposition (SVD) method, so as to finish splicing.
Drawings
Fig. 1 is a flowchart of a method for stitching three-dimensional point cloud data according to the present invention.
Figure 2 shows a flow diagram of one embodiment of the present invention.
FIG. 3 is a schematic structural diagram of rejecting mismatching point pairs according to an included angle between corresponding point normals.
Detailed Description
As shown in fig. 1, the method for stitching three-dimensional point cloud data includes the following steps:
(1) the point cloud acquisition system acquires partially overlapped source point cloud and target point cloud, and preprocesses the point cloud;
(2) the method comprises the steps of obtaining good initial positions of a source point cloud and a target point cloud through coordinate transformation of scanning space positions twice;
(3) calculating curvature characteristics of the point cloud, accelerating the searching process of nearest neighbor by using a Kdtree to find an initial matching point pair, and setting a constraint condition by using a direction vector threshold method to eliminate an error point pair;
(4) and obtaining a rotation matrix R and a translational vector t between the two point clouds by using a Singular Value Decomposition (SVD) method, and completing splicing.
Firstly, acquiring a source point cloud and a target point cloud which are partially overlapped through a point cloud acquisition system, and preprocessing the point clouds; then, the source point cloud and the target point cloud are enabled to obtain good initial positions through the coordinate transformation of scanning space positions twice; acquiring initial matching point pairs of two point clouds by calculating curvature characteristics of the point clouds and using a Kdtree nearest neighbor searching method, and setting constraint conditions by a direction vector threshold method to eliminate wrong point pairs; and finally, obtaining a rotation matrix R and a translational vector t between the two point clouds by utilizing singular value decomposition. Therefore, the problems of high splicing difficulty and low precision of two pieces of point cloud with low overlapping rate can be solved, and the splicing precision and the time efficiency of the point cloud data with low initial overlapping degree are effectively improved.
Preferably, in the step (1), the point cloud preprocessing includes filtering and down-sampling, and the noise and outliers of the original point cloud are removed by adopting methods of statistical filtering, radius filtering and gaussian filtering, so that the calculation efficiency and precision of subsequent operations are improved; the improved voxel filtering is adopted to carry out down-sampling on the point clouds, the number of the point clouds is reduced while the data structure of the original point clouds is not changed, and the speed of subsequent splicing is improved.
Preferably, in step (1), the VoxelGrid class of PCL is to create a three-dimensional voxel grid from the input point cloud data, and approximate the center of gravity of all points in the voxel to display other points in the voxel in each voxel, so that all points in the voxel are represented by a center of gravity point, which is not necessarily the point in the original point cloud, and there is a loss of the fine features of the original point cloud; the improved voxel filtering replaces the voxel gravity center point with the point closest to the voxel gravity center point in the original point cloud data so as to improve the expression accuracy of the point cloud data.
Preferably, the step (3) comprises the following substeps:
(3.1) constructing a covariance matrix through a query point x and k neighborhood points thereof in the point cloud by using the curvature characteristics:
wherein x isiIs the k neighborhood of the query point x,
is the centroid of x and its neighborhood points;
(3.2) obtaining three eigenvalues of the covariance matrix by a Singular Value Decomposition (SVD) method, and arranging the three eigenvalues from large to small, wherein the three eigenvalues are sequentially lambdad1>λd2>λd3；
(3.3) at query point xS for surface curvaturedRepresents:
(3.4) in order to enhance robustness, different neighborhood search radii are selected, and the difference of curvature information calculated under different search radii is used as matching information
Δsd＝sd+1-sd
(3.5) calculating the curvature information of another point cloud by using the same method, and recording the curvature information as Delta sd', setting a threshold value tau, and satisfying Δ sd-Δsd′<τ, considered as a pair of matching points;
(3.6) Direction vector thresholding method for Point p in Point cloudiAnd k neighborhood points of the local plane are fitted for the neighborhood points By using a least square method, the general expression of a plane equation is Ax + By + Cz-D which is 0, the actual meaning of vectors (A, B and C) of which the elements are not 0 at the same time in the four parameters to be solved is a normal vector n of the plane equation, and D is the distance from the original point to the plane; local plane P derived from least squareslThe fitting procedure is represented as:
by minimizing the objective function, the point of the vector formed by the point and each adjacent point thereof and the normal vector is multiplied by 0; in the above formula, m is a point xiAnd the centroids of its neighborhood points
At the same time make yi＝Xi-m
Thus, the objective function translates into:
(3.7) making S ═ YYT) Decomposing the mixture by using an SVD algorithm to obtain
Y＝U∑VT
The last column in U is the normal vector n to be solved, namely the eigenvector corresponding to the minimum eigenvalue;
and (3.8) unitizing the obtained normal vectors, calculating the cosine value of the included angle between the normal vectors of each matching point pair, setting a threshold value epsilon, and if the cosine value of the included angle cos theta is less than epsilon, considering the corresponding point pair as an error point pair and removing the error point pair from the corresponding point set.
Preferably, in the step (3.5), Kdtree is adopted to accelerate the nearest neighbor searching process when searching the neighborhood points of the query point.
Preferably, in the step (3.7), in the process of calculating the normal vector, an OpenMP multi-thread parallel computing method is adopted.
As shown in fig. 2, an embodiment of the present invention and its implementation process includes the following steps:
step 1: and acquiring partially overlapped source point cloud and target point cloud required by the experiment.
The 1 st group is a bunny partial data set, wherein source point clouds 4156 points and target point clouds 3679 points do not need to be preprocessed; the 2 nd group is point cloud data actually obtained by a structured light system, comprising 204833 points of source point cloud and 218594 points of target point cloud, wherein the two point clouds are down-sampled by improved voxel filtering, and the source point cloud is 51797 points and the target point cloud is 58216 points after pretreatment. Before point cloud preprocessing, the format of point cloud data acquired by structured light is txt, the format of the point cloud required by an experiment is pcd, the txt is converted into the pcd-format point cloud data under the environment that PCL 1.9.1 is configured by adopting software Visual Studio 2017, and the subsequent splicing process is also under the configuration environment.
Step 2: and determining an initial matching point pair of the two-point cloud by utilizing a coresponsence estimating normative shooting class.
The coresponsence-like optimization normal shooting realizes the calculation of corresponding point pairs of the input point cloud and the target point cloud under the constraint of the minimum distance through a normal line. The input is target and source point clouds, and the output is a point pair, namely, a corresponding point set between two groups of point clouds is output. It should be noted that the normal calculated here is the normal corresponding to the source point cloud, and not the normal corresponding to the target point cloud. When determining the corresponding relationship between the source point cloud and the target point cloud, there are two ways, one is degree correlation, the maximum distance max _ distance allowed between the source point cloud and the corresponding target point cloud is input, and the found corresponding relationship (the index of the query point, the index of the target point and the distance between the query point and the target point) is output and stored in the correlation; the other is the determinerecapalcorrespondances, which function is the same as above, but does not require the maximum distance allowed between the input source point cloud and the target point cloud where the corresponding points searched for are mutual. The initial matching point pairs determined by the group 1 point cloud are 2498 pairs and the initial matching point pairs determined by the group 2 point cloud are 22286 pairs.
And step 3: and eliminating wrong matching point pairs by using a direction vector threshold method.
The coresponsence rejection surface normal-like method realizes a corresponding relation removing method based on corresponding point normal angles. Inputting an included angle threshold value between the normal lines of the source point cloud and the target point cloud, the initial corresponding relation correspondents and the set corresponding point normal line, outputting an accurate matching point pair after the wrong corresponding relation is eliminated, and storing the accurate matching point pair in correspondents _ after _ projector. An OpenMP multithreading parallel acceleration method is adopted in the normal line calculation process, a Kdtree acceleration search method is adopted in the neighbor point search process, the calculation efficiency of the algorithm is improved, and the number of threads of OpenMP set in experiments is 8.
Fig. 3 is a schematic diagram of rejecting the pairs of mismatching points through the included angle between the normal lines of the corresponding points. Wherein q is1，q2，q3… … are partial points in the point cloud, n represents q1Normal vector at point, (q)1，q′1)，(q2，q′2)，(q3，q′3) Is the correct corresponding point pair. q. q.s4The normal vector at a point is n4，q′4The normal vector at a point is n'4，n4And n'4The included angle between the two is theta, a threshold value epsilon is set, if cos theta<ε, then (q)4，q′4) And the matching point pairs are wrong and are removed from the initial matching point pairs. In the experiment, the set included angle threshold value of the 1 st group of point clouds is 4 degrees, and the number of the rejected point pairs is 44 pairs; the set included angle threshold value of the 2 nd group of point clouds is 1 degree, and the number of the rejected point pairs is 127 pairs.
And 4, step 4: and obtaining a rotation matrix R and a translation vector t between the two point clouds by using a Singular Value Decomposition (SVD) method.
The transformational optimization SVD is the realization of an algorithm for estimating a transformation matrix for a given corresponding relation based on an SVD method, the input of the algorithm is two groups of point clouds needing to be registered and corresponding point pairs thereof, the output of the algorithm is the transformation matrix solved by the SVD method, and the transformation matrix is stored in the transformation.
The result data of two groups of point cloud data in the experiment are shown in table 1:
TABLE 1
In this embodiment, two sets of experimental data adopt bunny partial data sets and point cloud data actually acquired by a structured light system, and in the process of eliminating wrong matching point pairs by a direction vector threshold method, not only are matching point pairs reduced and subsequent calculation speed increased, but also wrong corresponding point pairs influencing splicing are eliminated, and the accuracy of the algorithm is improved.
It will be understood by those skilled in the art that all or part of the steps in the method of the above embodiments may be implemented by hardware instructions related to a program, the program may be stored in a computer-readable storage medium, and when executed, the program includes the steps of the method of the above embodiments, and the storage medium may be: ROM/RAM, magnetic disks, optical disks, memory cards, and the like. Therefore, corresponding to the method of the invention, the invention also comprises a three-dimensional point cloud data splicing device which is generally expressed in the form of functional modules corresponding to the steps of the method. The device includes:
the data preprocessing module is configured to acquire partially overlapped source point clouds and target point clouds through the point cloud acquisition system and preprocess the point clouds;
an initial position acquisition module configured to obtain good initial positions of the source point cloud and the target point cloud through coordinate transformation of scanning spatial positions twice;
the data eliminating module is configured to calculate the curvature characteristics of the point cloud, accelerate the searching process of nearest neighbor by using Kdtree to search for an initial matching point pair, and then set a constraint condition by using a direction vector threshold method to eliminate an error point pair;
and the splicing module is configured to obtain a rotation matrix R and a translational vector t between the two point clouds by utilizing a Singular Value Decomposition (SVD) method, so as to finish splicing.
The invention has the beneficial effects that: the invention provides a curvature characteristic-based Kdtree-accelerated nearest neighbor searching method aiming at the problems of high splicing difficulty, low precision and the like of two pieces of point clouds with low overlapping rate. After the matching point pairs of the two point clouds are obtained, in order to further reduce splicing errors, error point pairs are eliminated through a direction vector threshold method, and an OpenMP multi-thread parallel computing method is adopted in the process of computing a method vector, so that the algorithm efficiency is improved. And finally, obtaining a rotation matrix R and a translational vector t between the two point clouds by using a singular value decomposition method. The method effectively improves the splicing precision and time efficiency of the point cloud data with low initial overlapping degree.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and all simple modifications, equivalent variations and modifications made to the above embodiment according to the technical spirit of the present invention still belong to the protection scope of the technical solution of the present invention.
Claims (7)
1. A three-dimensional point cloud data splicing method is characterized by comprising the following steps: which comprises the following steps:
(1) the point cloud acquisition system acquires partially overlapped source point cloud and target point cloud, and preprocesses the point cloud;
(2) the method comprises the steps of obtaining good initial positions of a source point cloud and a target point cloud through coordinate transformation of scanning space positions twice;
(3) calculating curvature characteristics of the point cloud, accelerating the searching process of nearest neighbor by using a Kdtree to find an initial matching point pair, and setting a constraint condition by using a direction vector threshold method to eliminate an error point pair;
(4) and obtaining a rotation matrix R and a translational vector t between the two point clouds by using a Singular Value Decomposition (SVD) method, and completing splicing.
2. The method for stitching three-dimensional point cloud data according to claim 1, wherein: in the step (1), the point cloud preprocessing comprises filtering and down sampling, and original point cloud noise and outliers are removed by adopting methods of statistical filtering, radius filtering and Gaussian filtering, so that the calculation efficiency and precision of subsequent operation are improved; the improved voxel filtering is adopted to carry out down-sampling on the point clouds, the number of the point clouds is reduced while the data structure of the original point clouds is not changed, and the speed of subsequent splicing is improved.
3. The method for stitching three-dimensional point cloud data according to claim 2, wherein: in the step (1), a voxel grid class of the PCL is to create a three-dimensional voxel grid through the input point cloud data, and approximate the center of gravity of all points in the voxel to display other points in the voxel in each voxel, so that all points in the voxel are represented by a center of gravity point, which is not necessarily the point in the original point cloud, and there is a loss of the tiny characteristics of the original point cloud; the improved voxel filtering replaces the voxel gravity center point with the point closest to the voxel gravity center point in the original point cloud data so as to improve the expression accuracy of the point cloud data.
4. The method for stitching three-dimensional point cloud data according to claim 3, wherein: the step (3) comprises the following sub-steps:
(3.1) constructing a covariance matrix through a query point x and k neighborhood points thereof in the point cloud by using the curvature characteristics:
d＝1，2，...，
wherein x isiIs the k neighborhood of the query point x,
is the centroid of x and its neighborhood points;
(3.2) obtaining three eigenvalues of the covariance matrix by a Singular Value Decomposition (SVD) method, and arranging the three eigenvalues from large to small, wherein the three eigenvalues are sequentially lambdad1＞λd2＞λd3；
(3.3) surface curvature at query point x with sdRepresents:
(3.4) in order to enhance robustness, different neighborhood search radii are selected, and the difference of curvature information calculated under different search radii is used as matching information
Δsd＝sd+1-sd
(3.5) calculating the curvature information of another point cloud by using the same method, and recording the curvature information as Delta sd', setting a threshold value tau, and satisfying Δ sd-Δsd' < τ, considered as a matching point pair;
(3.6) Direction vector thresholding method for Point p in Point cloudiAnd k neighborhood points of the vector are obtained, a local plane is fitted for the neighborhood points By using a least square method, the general expression of a plane equation is Ax + By + Cz-D which is 0, and the actual meaning of vectors (A, B and C) of which the elements are not 0 at the same time in four parameters to be solved is thatThe normal vector n of the plane equation, and D is the distance from the origin to the plane; local plane P derived from least squareslThe fitting procedure is represented as:
by minimizing the objective function, the point of the vector formed by the point and each adjacent point thereof and the normal vector is multiplied by 0; in the above formula, m is a point xiAnd the centroids of its neighborhood points
At the same time make yi＝Xi-m
Thus, the objective function translates into:
(3.7) making S ═ YYT) Decomposing the mixture by using an SVD algorithm to obtain
Y＝U∑VT
The last column in U is the normal vector n to be solved, namely the eigenvector corresponding to the minimum eigenvalue;
and (3.8) unitizing the obtained normal vectors, calculating the cosine value of the included angle between the normal vectors of each matching point pair, setting a threshold value epsilon, and if the cosine value cos theta of the included angle is less than epsilon, considering the corresponding point pair as an error point pair and removing the error point pair from the corresponding point set.
5. The method for stitching three-dimensional point cloud data according to claim 4, wherein: in the step (3.5), Kdtree is adopted to accelerate the nearest neighbor searching process when searching the neighborhood points of the query points.
6. The method for stitching three-dimensional point cloud data according to claim 5, wherein: in the step (3.7), an OpenMP multithread parallel computing method is adopted in the process of computing the normal vector.
7. A three-dimensional point cloud data splicing device is characterized in that: it includes:
the data preprocessing module is configured to acquire partially overlapped source point clouds and target point clouds through the point cloud acquisition system and preprocess the point clouds;
an initial position acquisition module configured to obtain good initial positions of the source point cloud and the target point cloud through coordinate transformation of scanning spatial positions twice;
the data eliminating module is configured to calculate the curvature characteristics of the point cloud, accelerate the searching process of nearest neighbor by using Kdtree to search for an initial matching point pair, and then set a constraint condition by using a direction vector threshold method to eliminate an error point pair;
and the splicing module is configured to obtain a rotation matrix R and a translational vector t between the two point clouds by utilizing a Singular Value Decomposition (SVD) method, so as to finish splicing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110993313.1A CN113706381A (en) | 2021-08-26 | 2021-08-26 | Three-dimensional point cloud data splicing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110993313.1A CN113706381A (en) | 2021-08-26 | 2021-08-26 | Three-dimensional point cloud data splicing method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113706381A true CN113706381A (en) | 2021-11-26 |
Family
ID=78655734
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110993313.1A Pending CN113706381A (en) | 2021-08-26 | 2021-08-26 | Three-dimensional point cloud data splicing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113706381A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114677322A (en) * | 2021-12-30 | 2022-06-28 | 东北农业大学 | Milk cow body condition automatic scoring method based on attention-guided point cloud feature learning |
| CN116452648A (en) * | 2023-06-15 | 2023-07-18 | 武汉科技大学 | Point cloud registration method and system based on normal vector constraint correction |
| CN117961197A (en) * | 2024-04-01 | 2024-05-03 | 贵州大学 | Self-adaptive deviation rectifying method of unmanned turbine blade micropore electric machining unit |
-
2021
- 2021-08-26 CN CN202110993313.1A patent/CN113706381A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114677322A (en) * | 2021-12-30 | 2022-06-28 | 东北农业大学 | Milk cow body condition automatic scoring method based on attention-guided point cloud feature learning |
| CN116452648A (en) * | 2023-06-15 | 2023-07-18 | 武汉科技大学 | Point cloud registration method and system based on normal vector constraint correction |
| CN116452648B (en) * | 2023-06-15 | 2023-09-22 | 武汉科技大学 | Point cloud registration method and system based on normal vector constraint correction |
| CN117961197A (en) * | 2024-04-01 | 2024-05-03 | 贵州大学 | Self-adaptive deviation rectifying method of unmanned turbine blade micropore electric machining unit |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109166149B (en) | Positioning and three-dimensional line frame structure reconstruction method and system integrating binocular camera and IMU | |
| Zeng et al. | 3dmatch: Learning local geometric descriptors from rgb-d reconstructions | |
| CN113706381A (en) | Three-dimensional point cloud data splicing method and device | |
| CN106204574B (en) | Camera pose self-calibrating method based on objective plane motion feature | |
| CN112001955A (en) | Point cloud registration method and system based on two-dimensional projection plane matching constraint | |
| US11392787B2 (en) | Method for grasping texture-less metal parts based on bold image matching | |
| JP2009093611A (en) | System and method for recognizing three-dimensional object | |
| WO2009021959A1 (en) | Method and system for aligning three-dimensional surfaces | |
| CN114972459B (en) | Point cloud registration method based on low-dimensional point cloud local feature descriptor | |
| CN111986219B (en) | Matching method of three-dimensional point cloud and free-form surface model | |
| CN111145232A (en) | Three-dimensional point cloud automatic registration method based on characteristic information change degree | |
| CN111768447B (en) | Monocular camera object pose estimation method and system based on template matching | |
| CN111815686A (en) | Coarse-to-fine point cloud registration method based on geometric features | |
| CN112200915A (en) | Front and back deformation amount detection method based on target three-dimensional model texture image | |
| CN113313200B (en) | Point cloud precision matching method based on normal constraint | |
| CN111798453A (en) | Point cloud registration method and system for unmanned auxiliary positioning | |
| Chen et al. | Eyes localization algorithm based on prior MTCNN face detection | |
| CN111553410B (en) | Point cloud identification method based on key point local curved surface feature histogram and spatial relationship | |
| CN113393524A (en) | Target pose estimation method combining deep learning and contour point cloud reconstruction | |
| JPH07103715A (en) | Method and apparatus for recognizing three-dimensional position and attitude based on visual sense | |
| Liu et al. | A fast weighted registration method of 3d point cloud based on curvature feature | |
| Guo et al. | A fast global method combined with local features for 6d object pose estimation | |
| CN113256693A (en) | Multi-view registration method based on K-means and normal distribution transformation | |
| Kang et al. | 3D urban reconstruction from wide area aerial surveillance video | |
| CN106709942A (en) | Panoramic image mistaken matching elimination method based on characteristic azimuth |
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 |