CN112884901B - Three-dimensional point cloud data normal global consistency method for semi-closed space scene - Google Patents
Three-dimensional point cloud data normal global consistency method for semi-closed space scene Download PDFInfo
- Publication number
- CN112884901B CN112884901B CN202110198577.8A CN202110198577A CN112884901B CN 112884901 B CN112884901 B CN 112884901B CN 202110198577 A CN202110198577 A CN 202110198577A CN 112884901 B CN112884901 B CN 112884901B
- Authority
- CN
- China
- Prior art keywords
- normal
- point cloud
- voxel
- cloud data
- plane
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a three-dimensional point cloud data normal global uniformization method of a semi-closed space scene, which comprises the steps of firstly calculating the normal direction of three-dimensional point cloud data when the three-dimensional point cloud data is obtained, counting the normal distribution to obtain a point cloud main direction, then rotating the point cloud to the main direction, performing voxelization on the space where the point cloud is located, calculating the normal direction of a voxel, designing a normal global uniformization energy optimization equation according to the matching result of the voxel in the main direction, solving the equation to obtain the normal direction of the voxel to further obtain the normal direction of an inner point of the voxel, further optimizing the normal direction of the inner point by adopting an in-plane point voting strategy, and finally obtaining the point cloud with the globally uniform normal direction. The invention overcomes the local ambiguity characteristics of the point cloud data in the conventional normal calculation, provides the working process of automatically, quickly and globally unifying the normal direction of the point cloud, can process the three-dimensional point cloud data of different sources, and provides a new solution for applying the three-dimensional point cloud data to feature extraction, registration, surface reconstruction and visualization.
Description
Technical Field
The invention belongs to the technical field of three-dimensional point cloud data processing, relates to a point cloud data normal global consistency method, and particularly relates to a three-dimensional point cloud data normal vector direction consistency method for a semi-closed space scene.
Background
The point cloud (X, Y, Z, A) becomes a third important space-time data source following the vector map and the image data, has incomparable superiority with the two-dimensional vector map and the image, is a main source for acquiring three-dimensional geographic information, and has irreplaceable important function on the fine description of the three-dimensional space. With the rapid development of related technologies such as sensor technology and chip technology, devices and technologies such as three-dimensional laser scanners, multi-view (dense) matching, depth cameras, mobile measurement systems integrating laser-attitude sensors and cameras, and the like, can acquire a large amount of three-dimensional coordinate information of a target surface in a three-dimensional point cloud manner at high speed, high density and high precision, and are generally called point cloud data. Due to the difference of physical characteristics of the acquisition modes, the point cloud data has the problems of measurement error, uneven density distribution, incomplete data caused by object shielding, only containing geometric characteristics and lacking semantic information and the like, and the point cloud needs to be preprocessed to improve the quality of the point cloud. In addition to basic three-dimensional point position information, the most important information on the three-dimensional point cloud is the normal direction of the three-dimensional point, other associated geometric information (curvature, characteristic line and the like) can be obtained through the position and normal direction calculation of the point, and the accuracy of the position and normal direction of the point directly influences the precision and effect of related applications. However, the currently widely used method of point cloud normal vector calculation is obtained by function fitting of local neighborhood, including linear function and high-order polynomial function. The directions of the normal vectors obtained by the fitting functions are ambiguous, that is, the directions of the normal vectors and the opposite directions of the normal vectors are locally correct, but the directions of the normal directions of the cloud are discontinuous and inconsistent in overall view. Therefore, a normal vector with a globally consistent direction is a basic requirement for point clouds for geometric processing.
Disclosure of Invention
The invention aims to search a normal vector orientation method for automatically generating a global consistent direction by using three-dimensional point cloud data, research the characteristics of the three-dimensional point cloud in order to solve the problems, provide a scientific and reasonable global consistent orientation method, automatically determine the normal direction and provide a new solution for applying the three-dimensional point cloud data to feature extraction, registration, surface reconstruction and visualization.
The technical scheme adopted by the invention is as follows: a three-dimensional point cloud data normal global consistency method for a semi-closed space scene comprises the following steps:
step 1: calculating the main direction of the point cloud data;
step 2: calculating the normal global consistent direction of the point cloud data;
the specific implementation comprises the following substeps:
step 2.1: voxelizing the three-dimensional space occupied by the point cloud data according to the size of the input voxel, neglecting the voxel without points in the voxel, and calculating the normal direction of each voxel by the points in the voxel;
step 2.2: pairing the voxels according to the normal principal direction;
step 2.3: constructing a normal global uniformization energy optimization equation according to the paired normal main directions;
the normal global uniformization energy optimization equation Elabel(L) is:
where L { -1, +1}, the marking direction L ultimately determined for each nodeiThe normal vector of the voxel corresponding to the node is ni←Li·ni,p∈P,pm∈P,(p,pm)∈pairi(i-0, 1,2) indicates that the voxel nodes P and pm form a pair in the normal i principal direction, i is a mark in the normal principal direction (X-0, Y-1, Z-2), the voxel set P, λ is a constant, (P, q) ∈ epsilon indicates two neighboring voxels, ω is a constant, and m is a constant∠p,qThe absolute value of the cosine of the normal vector included angle of the adjacent voxels is obtained; ρ (L)p≠Lq) Represents the Potts model when Lp≠LqWhen the rho value is equal to 0, otherwise, the rho value is equal to 1; n isp,iRepresenting the component of node p in the normal i direction;
step 2.4: solving an energy minimization equation by adopting an optimization equation to obtain a normal vector direction of each voxel;
step 2.5: and (3) endowing the normal direction of the voxel to the normal direction of the points contained in the voxel so as to obtain the normal vector direction of the point cloud.
The invention can process point cloud data of different sources, including point cloud data of single-station or multi-station registration and splicing of a ground three-dimensional laser scanner, point cloud data obtained by multi-view image geometric calculation, point cloud data obtained by a depth camera and an indoor mobile measurement system integrating a laser-attitude sensor and a camera, and the like.
The invention has the following beneficial effects:
the invention provides an automatic optimization process of three-dimensional point cloud normal global consistency, and provides a new method for point cloud data processing.
The invention comprises a normal pairing method, which can effectively process indoor scene point cloud.
3, the method can process three-dimensional point cloud data from different sources, and particularly has a better effect on scenes with the characteristic of closure, such as three-dimensional point clouds obtained in indoor scenes.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention.
Fig. 2 is a schematic view of voxel pairing according to an embodiment of the present invention.
Detailed Description
In order to facilitate the understanding and implementation of the present invention for those of ordinary skill in the art, the present invention is further described in detail with reference to the accompanying drawings and examples, it is to be understood that the embodiments described herein are merely illustrative and explanatory of the present invention and are not restrictive thereof.
Referring to fig. 1, the method for normal global unification of three-dimensional point cloud data of a semi-closed space scene provided by the invention comprises the following steps:
step 1: calculating the main direction of the point cloud data;
in this embodiment, the specific implementation of step 1 includes the following substeps:
step 1.1: calculating the normal direction of point cloud data (X, Y, Z, A) and carrying out plane clustering; wherein X, Y, Z represent the three-dimensional coordinates of the point;
step 1.2: dividing a unit hemisphere into n parts (the value of n is larger, and the value of n is 1000 in the embodiment);
step 1.3: taking the number of points in the plane as the weighting frequency, and counting a direction histogram of the normal direction of the plane;
step 1.4: setting the normal average value of the plane in the maximum frequency group in the histogram as a first main direction, finding the group with the maximum frequency in the range of 80-100 degrees away from the direction, taking the normal average value of the plane in the direction in the group as a second main direction, and similarly finding a third main direction in the range of 80-100 degrees away from the first main direction and the second main direction;
step 1.5: rotating the three-dimensional point cloud to the calculated main direction according to the main direction and the formula (1);
in the formula, T4×4The transformation matrix is 4 multiplied by 4 and is obtained by calculating three rotation angles in the step 1.4; the position and normal vector of the point cloud are calculated by the formula (1), T is 4 multiplied by 4 vector, P is 4 multiplied by 1 vector, P' is 4 multiplied by 1 vector, when P is xyz space coordinate, the formula (1) T is directly used4×4Multiplying the translation and rotation part vectors to obtain a transformed position P'; similarly, when P is the normal vector nxnynz, T is used4×4The rotated components of (a) are multiplied to obtain a rotated normal vector P', where there is no translation.
Step 2: calculating the normal global consistent direction of the point cloud data;
in this embodiment, the specific implementation of step 2 includes the following substeps:
step 2.1: voxelizing the three-dimensional space occupied by the point cloud data according to the size of the input voxel, neglecting the voxel without points in the voxel, and calculating the normal direction of each voxel by the points in the voxel;
step 2.2: please refer to fig. 2, the voxels are paired according to the normal principal direction;
step 2.3: constructing a normal global uniformization energy optimization equation according to the paired normal main directions;
wherein, the normal global uniformization energy optimization equation Elabel(L) is represented by the following formula:
where L { -1, +1}, the marking direction L ultimately determined for each nodeiThe normal vector of the voxel corresponding to the node is ni←Li·ni,p∈P,pm∈P,(p,pm)∈pairi(i-0, 1,2) indicates that the voxel nodes P and pm form a pair in the normal i principal direction, i is a mark in the normal principal direction (X-0, Y-1, Z-2), the voxel set P, λ is a constant, (P, q) ∈ epsilon indicates two neighboring voxels, ω is a constant, and m is a constant∠p,qThe absolute value of the cosine of the normal vector included angle of the adjacent voxels is obtained; ρ (L)p≠Lq) Represents the Potts model when Lp≠LqWhen the rho value is equal to 0, otherwise, the rho value is equal to 1; n isp,iRepresenting the component of node p in the normal i direction;
step 2.4: solving an energy minimization equation (such as linear programming or graph cut optimization and the like) by adopting an optimization equation to obtain the normal vector direction of each voxel;
step 2.5: and (3) endowing the normal direction of the voxel to the normal direction of the points contained in the voxel so as to obtain the normal vector direction of the point cloud.
And step 3: optimizing the normal direction of the points in the plane by adopting a voting rule;
in this embodiment, the specific implementation of step 3 includes the following substeps:
step 3.1: counting the normal directions of the points in the plane, and comparing the number of the points which are consistent with the normal direction of the plane with the number of the points which are inconsistent with the normal direction of the plane;
step 3.2: determining a plane direction from a direction having a larger number;
step 3.3: updating all directions of the interior points by using the plane direction;
the updating method of the embodiment comprises the following steps: if the direction of the inner point is not consistent with the direction of the plane, namely the dot product of the normal vector is negative, multiplying the normal vector of the inner point by-1, or directly endowing the normal direction of the plane to the inner point;
step 3.4: using T in formula (1)4×4The inverse matrix of (A) rotates the point cloudAnd giving a new normal direction to the original point cloud according to the original direction or the point cloud ID number to finally obtain the point cloud with the globally consistent normal direction.
In specific implementation, those skilled in the art can support the implementation process by using a computer software mode.
It should be understood that the above description of the preferred embodiments is given for clarity and not for any purpose of limitation, and that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (3)
1. A three-dimensional point cloud data normal global consistency method for a semi-closed space scene is characterized by comprising the following steps:
step 1: calculating the main direction of the point cloud data;
step 2: calculating the normal global consistent direction of the point cloud data;
the specific implementation comprises the following substeps:
step 2.1: voxelizing the three-dimensional space occupied by the point cloud data according to the size of the input voxel, neglecting the voxel without points in the voxel, and calculating the normal direction of each voxel by the points in the voxel;
step 2.2: pairing the voxels according to the normal principal direction;
step 2.3: constructing a normal global uniformization energy optimization equation according to the paired normal main directions;
the normal global uniformization energy optimization equation Elabel(L) is:
where L { -1, +1}, the marking direction L ultimately determined for each nodeiThe normal vector of the voxel corresponding to the node is ni←Li·ni,p∈P,pm∈P,(p,pm)∈pairiIndicating that voxel nodes p and pm form a pair in the normal principal direction i, i being the normal principal direction (X, Y, Z) index, i being 0,1,2, X being 0, Y being 1, Z being 2; the voxel set P, lambda is a constant, (P, q) epsilon represents two adjacent voxels, omega∠p,qThe absolute value of the cosine of the normal vector included angle of the adjacent voxels is obtained; ρ represents the Potts model when Lp≠LqWhen the rho value is equal to 0, otherwise, the rho value is equal to 1; n isp,iRepresenting the component of node p in the normal i direction;
step 2.4: solving an energy minimization equation by adopting an optimization equation to obtain a normal vector direction of each voxel;
step 2.5: and (3) endowing the normal direction of the voxel to the normal direction of the points contained in the voxel so as to obtain the normal vector direction of the point cloud.
2. The method for the normal global consistency of the three-dimensional point cloud data of the semi-closed space scene according to claim 1, wherein the specific implementation of the step 1 comprises the following sub-steps:
step 1.1: calculating the normal direction of point cloud data (X, Y, Z) and carrying out plane clustering; wherein X, Y, Z represent the three-dimensional coordinates of the point;
step 1.2: dividing a unit hemisphere into n parts;
step 1.3: taking the number of points in the plane as the weighting frequency, and counting a direction histogram of the normal direction of the plane;
step 1.4: setting the normal average value of the plane in the maximum frequency group in the histogram as a first main direction, finding the group with the maximum frequency in the range of 80-100 degrees away from the direction, taking the normal average value of the plane in the direction in the group as a second main direction, and similarly finding a third main direction in the range of 80-100 degrees away from the first main direction and the second main direction;
step 1.5: rotating the three-dimensional point cloud to the calculated main direction according to the main direction and the formula (1);
in the formula, T4×4The transformation matrix is 4 multiplied by 4 and is obtained by calculating three rotation angles in the step 1.4; the position and normal vector of the point cloud are calculated by formula (1), T is 4 × 4 vector, P is 4 × 1 vector, P' is 4 × 1 vector, when P is space coordinate (x, y, z,1)TWhen, directly using the formula (1) T4×4Multiplying the translation and rotation part vectors to obtain a transformed position P'; similarly, when P is a normal vector (nx, ny, nz,1)TWhen using T4×4The rotated components of (a) are multiplied to obtain a rotated normal vector P', where there is no translation.
3. The method for the normal global unification of the three-dimensional point cloud data of the semi-closed space scene according to any one of claims 1-2, further comprising the step 3: optimizing the normal direction of a point in a plane; the specific implementation comprises the following substeps:
step 3.1: counting the normal directions of the points in the plane, and comparing the number of the points which are consistent with the normal direction of the plane with the number of the points which are inconsistent with the normal direction of the plane;
step 3.2: determining a plane direction from a direction having a larger number;
step 3.3: updating all directions of the interior points by using the plane direction;
the updating method comprises the following steps: if the direction of the inner point is not consistent with the direction of the plane, namely the dot product of the normal vector is negative, multiplying the normal vector of the inner point by-1, or directly endowing the normal direction of the plane to the inner point;
step 3.4: and rotating the point cloud to the original direction or endowing the original point cloud with a new normal direction according to the point cloud ID number to finally obtain the point cloud with the globally consistent normal direction.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110198577.8A CN112884901B (en) | 2021-02-22 | 2021-02-22 | Three-dimensional point cloud data normal global consistency method for semi-closed space scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110198577.8A CN112884901B (en) | 2021-02-22 | 2021-02-22 | Three-dimensional point cloud data normal global consistency method for semi-closed space scene |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112884901A CN112884901A (en) | 2021-06-01 |
CN112884901B true CN112884901B (en) | 2022-04-15 |
Family
ID=76056862
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110198577.8A Active CN112884901B (en) | 2021-02-22 | 2021-02-22 | Three-dimensional point cloud data normal global consistency method for semi-closed space scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112884901B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105046688A (en) * | 2015-06-23 | 2015-11-11 | 北京工业大学 | Method for automatically identifying multiple planes in three-dimensional point cloud |
WO2016207669A2 (en) * | 2015-06-23 | 2016-12-29 | Oxford University Innovation Limited | A method of generating a three dimensional representation of an environment or system |
CN106803267A (en) * | 2017-01-10 | 2017-06-06 | 西安电子科技大学 | Indoor scene three-dimensional rebuilding method based on Kinect |
CN109887082A (en) * | 2019-01-22 | 2019-06-14 | 武汉大学 | A kind of interior architecture three-dimensional modeling method and device based on point cloud data |
CN110222642A (en) * | 2019-06-06 | 2019-09-10 | 上海黑塞智能科技有限公司 | A kind of planar architectural component point cloud contour extraction method based on global figure cluster |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109325998B (en) * | 2018-10-08 | 2023-06-30 | 香港理工大学 | Indoor 3D modeling method, system and related device based on point cloud data |
-
2021
- 2021-02-22 CN CN202110198577.8A patent/CN112884901B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105046688A (en) * | 2015-06-23 | 2015-11-11 | 北京工业大学 | Method for automatically identifying multiple planes in three-dimensional point cloud |
WO2016207669A2 (en) * | 2015-06-23 | 2016-12-29 | Oxford University Innovation Limited | A method of generating a three dimensional representation of an environment or system |
CN106803267A (en) * | 2017-01-10 | 2017-06-06 | 西安电子科技大学 | Indoor scene three-dimensional rebuilding method based on Kinect |
CN109887082A (en) * | 2019-01-22 | 2019-06-14 | 武汉大学 | A kind of interior architecture three-dimensional modeling method and device based on point cloud data |
CN110222642A (en) * | 2019-06-06 | 2019-09-10 | 上海黑塞智能科技有限公司 | A kind of planar architectural component point cloud contour extraction method based on global figure cluster |
Non-Patent Citations (1)
Title |
---|
法向估计的屏蔽泊松算法三维点云重建;欧阳宁等;《电视技术》;20171231;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN112884901A (en) | 2021-06-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9117267B2 (en) | Systems and methods for marking images for three-dimensional image generation | |
Stahs et al. | Fast and robust range data acquisition in a low-cost environment | |
CN104019799A (en) | Relative orientation method by using optimization of local parameter to calculate basis matrix | |
GB2566443A (en) | Cross-source point cloud registration | |
Cheng et al. | Extracting three-dimensional (3D) spatial information from sequential oblique unmanned aerial system (UAS) imagery for digital surface modeling | |
Knyaz et al. | Accuracy evaluation of structure from motion surface 3D reconstruction | |
Xu et al. | 3D reconstruction method based on second-order semiglobal stereo matching and fast point positioning Delaunay triangulation | |
Jabborov et al. | Image-based camera localization algorithm for smartphone cameras based on reference objects | |
Yilmazturk et al. | Geometric evaluation of mobile‐phone camera images for 3D information | |
Huang et al. | Registration method for terrestrial LiDAR point clouds using geometric features | |
CN112884901B (en) | Three-dimensional point cloud data normal global consistency method for semi-closed space scene | |
CN117522963A (en) | Corner positioning method and device of checkerboard, storage medium and electronic equipment | |
Wu | Photogrammetry for 3D mapping in Urban Areas | |
Liu et al. | Camera calibration method with checkerboard pattern under complicated illumination | |
Feng et al. | A closed-form single-pose calibration method for the camera–projector system | |
Li et al. | BDLoc: Global localization from 2.5 D building map | |
Zhou et al. | A novel method for reconstructing general 3D curves from stereo images | |
Yu et al. | Multi-view 2D–3D alignment with hybrid bundle adjustment for visual metrology | |
Xu et al. | Automatic registration method for TLS LiDAR data and image-based reconstructed data | |
Yang et al. | Three-dimensional panoramic terrain reconstruction from aerial imagery | |
Kumar et al. | To Study the Effect of Varying Camera Parameters in the Process of Matching and Reconstruction | |
Arslan | Accuracy assessment of single viewing techniques for metric measurements on single images | |
Yu et al. | A spatial self-calibration algorithm for robot vision | |
Costantino et al. | ICV and fine-registration algorithms for an efficient merging of point clouds | |
Zhou et al. | Digitization of cultural heritage |
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 |