CN105678833A - Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling - Google Patents
Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling Download PDFInfo
- Publication number
- CN105678833A CN105678833A CN201610015916.3A CN201610015916A CN105678833A CN 105678833 A CN105678833 A CN 105678833A CN 201610015916 A CN201610015916 A CN 201610015916A CN 105678833 A CN105678833 A CN 105678833A
- Authority
- CN
- China
- Prior art keywords
- relation
- geometric data
- view
- cloud geometric
- point cloud
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
Landscapes
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a point cloud geometrical data automatic splicing algorithm based on multiview image three-dimensional modeling. The algorithm comprises the following steps: step 1, in different views, calculating point cloud geometrical data of a three-dimensional object in need of modeling, and building feature matching point pairs of the point cloud geometrical data; step 2, picking a view randomly as a reference view, and using the feature matching point pairs of the point cloud geometrical data to calculate a corresponding relationship matrix representing the relative positions of all other views to the reference view; step 3, carrying out singular value decomposition of the corresponding relationship matrix, and calculating a relationship translation vector and a relationship rotation vector of the feature matching point pair of each view and the reference view; and step 4, on the basis of the relationship translation vector and the relationship rotation vector, displaying the point cloud geometrical data of each view in a reference view coordinate system, and completing automatic splicing of the point cloud geometrical data. The algorithm is simple and reliable, easy to implement, convenient for operation, and high in modeling precision.
Description
Technical field
The present invention relates to microcomputer modelling technical field, be specifically related to the automatic Mosaic algorithm of a kind of some cloud geometric data based on multi-view image three-dimensional modeling.
Background technology
In the process in multi-view image three-dimensional geometric mode building system, three-dimensional profile built, Geometric Modeling between adjacent viewpoint picture is merely capable of obtaining the geometric point cloud geometric data of testee surface regional area, want to obtain whole three-dimensional profile data to need to carry out multiple views acquisition image, being simultaneous for each two adjacent viewpoint and carry out Geometric Modeling, this causes that the geometric coordinate system of calculated some cloud geometric data under different points of view is different.
In order to obtain the whole three-dimensional profile geometric data on testee surface, it is necessary to the local geometric data under different coordinates are transformed under same unified coordinate system. Automatic Mosaic and the registration of the three-dimensional point cloud geometric data under the different coordinates that the modeling of multiple adjacent viewpoint picture obtains are always up a stubborn problem. Existing method mainly includes following several:
1) by the labelling point in measured object exterior pasting auxiliary, the labelling point of different measuring several times is scanned for building the labelling point pair of coupling, ensure the common labelling point having at least more than three between two viewpoints simultaneously, then pass through the common labelling point of these couplings, calculate the coordinate conversion relation between the some cloud geometric data that repeatedly different points of view measurement obtains, thus realizing the automatic Mosaic of multiple views point cloud geometric data.
But not only can destroy the texture information of measured object surface at object being measured surface mount aid mark point, object being measured surface geometry and data texturing that labelling point cannot be pasted covering place are modeled simultaneously. In addition the method is not suitable at the upper binding mark point of some special object being measured (such as history relic surface), and therefore it uses that to have certain scope restricted.
2) utilize some The Cloud Terraces to determine the change in location relation between object being measured and multiple views, directly calculated the changes in coordinates relation putting between cloud geometric data under multiple views by the kinematic parameter of The Cloud Terrace.
The method is reliable and stable, and has higher precision, but needs additional high-accuracy mechanical equipment, thus it is complicated to cause that multiple views obtains device structure, it is impossible to bigger object measurement.
3) the manual characteristic point choosing coupling is mated in advance, then passes through existing business software Processing Algorithm and completes a splicing for cloud geometric data.
First this type of method needs by the artificial coupling intervened in advance and realize data, but artificial matching error is excessive is unable to reach desirable splicing effect, it is impossible to realize the full-automatic splicing to the some cloud geometric data after multi-view image three-dimensional modeling.
Summary of the invention
The invention provides the automatic Mosaic algorithm of a kind of some cloud geometric data based on multi-view image three-dimensional modeling, without assisting equipment by mechanical hardware, aid mark point is pasted also without the three-dimensional object surface in modeling, automatic Mosaic and the coupling of the some cloud geometric data of multi-view image three-dimensional modeling can be completed, simple and reliable, it is easily achieved, using the teaching of the invention it is possible to provide higher modeling accuracy, there is wide applicability and practicality.
A kind of automatic Mosaic algorithm of the some cloud geometric data based on multi-view image three-dimensional modeling, including:
Step 1, under different points of view, calculates the some cloud geometric data of the three-dimensional body needing modeling, and builds the characteristic matching point pair of a cloud geometric data;
Step 2, randomly selects a viewpoint as reference view, utilizes the relation homography of the characteristic matching point pair of some cloud geometric data, each viewpoint of computational representation and reference view relative position relation;
Step 3, carries out singular value decomposition to relation homography, asks for relation translation vector and the relation rotating vector of characteristic matching point pair between each viewpoint (not including reference view) and reference view;
Step 4, according to relation translation vector and relation rotating vector, is indicated the some cloud geometric data in each viewpoint under reference view coordinate system, completes an automatic Mosaic for cloud geometric data.
The number of different points of view is more many, and automatic Mosaic and the some cloud geometric data obtained after coupling are more accurate, but corresponding amount of calculation also can greatly increase, it is preferable that the number of different points of view is at least 6.
Different points of view randomly selects a viewpoint as reference view, the some cloud geometric data under other viewpoints is converted to and represents under reference view coordinate system.
After relation homography is carried out singular value decomposition by step 3, relation translation vector and the relation rotating vector of characteristic matching point pair, the relation translation vector of characteristic matching point pair and relation rotating vector can be obtained namely had relation translation vector and the relation rotating vector of a cloud geometric data.
Point cloud geometric data under different points of view can be converted to by relation translation vector and relation rotating vector and represent under the coordinate system of reference view.
As preferably, the number of characteristic matching point pair is 100～120. It is preferred that, the number of characteristic matching point pair is 100.
When calculated relationship homography, the characteristic matching point centering putting cloud geometric data of other viewpoints except reference view with reference view randomly select 100 characteristic matching points to being calculated. (each viewpoint choose 100 characteristic matching points to)
N different viewpoint is numbered respectively, it is followed successively by 1,2,3......n, reference view, i.e. n=1, when calculating the relative geometrical relation between kth (k=2,3......n) individual viewpoint and reference view, from kth (k=2,3 ... n) individual viewpoint randomly selects 100 characteristic matching points pair with the characteristic matching point centering putting cloud geometric data of reference view, is calculated.
Kth (k=2,3 ... n) individual viewpoint collectively forms relation homography M with the relative geometrical relation of reference view.
As preferably, described step 2 using during calculated relationship homography optimization mechanism. Use optimization mechanism to may further ensure that the robustness of calculated relation homography, increase the fault-tolerance to characteristic matching point pair by mistake.
Assume that optimization mechanism repeatedly randomly selects in characteristic matching point centering, choose hundred pairs of characteristic matching points pair every time, for 100 characteristic matching points in the image 1 in a certain viewpoint, for these 100 characteristic matching point p(i=1. . . . 100), by epipolar geometry constraints, find its polar curve L corresponding in this visual point image 2(i=1. . . . 100), then calculate p(i=1. . . . 20)The corresponding characteristic point in this visual point image 2 is to L(i=1. . . . 100)Distance D(i=1. . . . 100), and calculate total distance D=D1+D2+D3+……+D99+D100, finally choose one group of minimum matching double points of total distance D value as final characteristic matching point pair. Adopt this optimization method can pass through limit geometrical constraint on the one hand and improve the fault-tolerance of characteristic matching point pair, also be able to ensure the robustness of algorithm on the other hand.
Relation homography M is carried out singular value decomposition by described step 3, singular value SVD decomposes (SingularValueDecomposition) can calculate the normalized relation translation vector T between two different points of view and relation spin matrix R, then utilize the relation translation vector T and relation spin matrix R of characteristic matching point pair between other each viewpoint and the reference view in multiple views, obtain the actual relationship translation vector T ' of some cloud geometric data relative reference viewpoint corresponding to each viewpoint.
The present invention is based on the automatic Mosaic algorithm of the some cloud geometric data of multi-view image three-dimensional modeling, only need to utilize the characteristic matching point pair under different points of view, the automatic Mosaic of the some cloud geometric data based on multi-view image three-dimensional modeling can be realized, simple and reliable, it is easily achieved, easy to operate, and higher modeling accuracy can be reached.
Accompanying drawing explanation
Fig. 1 is the present invention flow chart based on the automatic Mosaic algorithm of the some cloud geometric data of multi-view image three-dimensional modeling.
Detailed description of the invention
Below in conjunction with accompanying drawing, the present invention is described in detail.
As it is shown in figure 1, the automatic Mosaic algorithm of a kind of some cloud geometric data based on multi-view image three-dimensional modeling, comprise the steps:
(1) calculated relationship homography M
Under different points of view, shooting obtains being modeled the multiple image of object, and different points of view is labeled as 1,2,3 successively ... n, and randomly selecting a viewpoint is reference view, for instance taking n=1 is reference view.
Set up kth (k=2,3 ... n) random image I between individual viewpoint and reference viewkWith random image I1Between stable characteristic matching point pair, it is assumed that random image IkWith random image I1In the image coordinate respectively I corresponding under the camera coordinates system of respective viewpoint of characteristic pointkAnd I1, utilize three-dimensional vector to be expressed as (I1 k, I2 k, I3 k), (I1 1, I2 1, I3 1)。
Limit restraint equation can be obtained according to geometrical-restriction relation
(I1)TFIk=0 (1)
Wherein: F is fundamental matrix, it is a kind of Algebraic Expression of epipolar geom etry, is also a highly important matrix in multiple views three-dimensional modeling.
Meanwhile, F also meets following relationship
F=K2 -TEK1 -1(2)
K1And K2The respectively upper triangular matrix of 3 × 3, represents the inner parameter of two video cameras respectively, and E is matrix, contains the structural parameters between adjacent two viewpoints.
Formula (2) is brought in formula (1) following formula can be obtained
(I1)TK1 -TFK1 -1Ik=0 (3)
Assume random image IkWith random image I1In the normalization corresponding under the camera coordinates system of each measurement viewpoint of characteristic point after three-dimensional homogeneous image coordinate respectively Il kAnd Il 1, order
Ql k=K1 -1Ik(4)
Il 1=K1 -1I1(5)
Then epipolar-line constraint equation can be reduced to
(Il 1)TEQl k=0 (6)
Fundamental matrix F is usually the non-zero matrix of 3 × 3, and the value of its determinant is 0, namely
Det (F)=0 (7)
According to formula (2) it can be seen that matrix E also meets formula (7), matrix E has following character simultaneously
Utilize formula (7) and (8), by 100 algorithms, i.e. random image I between kth viewpoint and reference viewkWith random image I1Between choose 100 pairs of matched pixel points and try to achieve matrix E.
Utilizing 100 algorithm calculated relationship homography M, 100 algorithms are the iterative evaluation methods for calculating the geometric coordinate transformation relation between different points of view, and key step is as follows:
From random image IkWith random image I1Arbitrarily choose 100 groups in the invariant feature matching double points set set up, then this 100 stack features matching double points all meets formula (6), and therefore, epipolar-line constraint equation is represented by again
Wherein:
Pile up hundred vectors to characteristic matching point pair100 × 9 relation homography M can be obtained.
After asking for the kernel of relation homography M, the expansion of computing formula (7) and formula (8) respectively.
(2) the singular value SVD adopting matrix decompose (referring to wearing China. matrix theory. Beijing, Science Press, 2001) method relation homography M is carried out singular value matrix decomposition, obtain relation spin matrix R and the value of relation translation vector T.
The point cloud geometric data assumed under the geometric coordinate system that reference view is set up is X={Xi, i=1,2 ... }, kth (k=2,3 ... n) individual viewpoint set up geometric coordinate system under some cloud geometric data be X '={ X 'i, j=1,2 ... }.
In order to obtain entirety some cloud geometric data, to kth (k=2,3 ... ... n) the some cloud geometric data of individual viewpoint pass through geometric transformation coordinate, be transformed to and utilize the unified geometric coordinate system of reference view to represent.
Assume by kth (k=2,3 ... n) the some cloud geometric data of individual viewpoint converts through geometric coordinate, utilizes the some cloud geometric data that the coordinate system of reference view obtains after representing to beThen any one geometric data point X ' in some cloud geometric data set X 'iCoordinate transform formula be
Wherein: R represents kth (k=2,3 ... n) geometric coordinate of individual viewpoint is tied to the relation spin matrix of the geometric coordinate system of reference view;
T represents kth (k=2,3 ... n) geometric coordinate of individual viewpoint is tied to the relation translation vector of the geometric coordinate system of reference view.
Realize the some splicing of cloud geometric data and the coupling of different points of view, it is necessary to calculate the relation spin matrix R and relation translation vector T of two viewpoint geometric coordinate systems.
Utilize and represent that between two viewpoints, the relation homography M and relation homography M of relative geometry position are with the relation between relation spin matrix R and relation translation vector T, it is possible to obtain relation spin matrix R and relation translation vector T.
Relation between relation homography M, relation spin matrix R and relation translation vector T is as follows
Wherein, T=(t1, t2, t3)(14)
When calculating relation homography M, relation homography M is carried out Singular Value Decomposition Using and can obtain relation spin matrix R and the value of relation translation vector T.
(3) the relation translation vector T and relation spin matrix R of characteristic matching point pair between each viewpoint and reference view are utilized, the actual relationship translation vector T ', actual relationship translation vector T ' that put cloud geometric data relative reference viewpoint in each viewpoint can be calculated same with relation translation vector T-phase.
(4) according to relation spin matrix R and relation actual translation vector T ', point cloud geometric data under each viewpoint is carried out geometric coordinate conversion, utilize formula (15) will to be had cloud geometric data unified representation under reference view coordinate system, it is achieved under different points of view, to put automatic Mosaic and the coupling of cloud geometric data.
X=RX '+T ' (15)
Wherein, X is the some cloud geometric data under the geometric coordinate system that reference view is set up;
X ' is kth (k=2,3 ... n) the some cloud geometric data under the geometric coordinate system that individual viewpoint is set up;
R is relation spin matrix;
T ' is actual relationship translation vector.
Claims (5)
1. the automatic Mosaic algorithm based on the some cloud geometric data of multi-view image three-dimensional modeling, it is characterised in that including:
Step 1, under different points of view, calculates the some cloud geometric data of the three-dimensional body needing modeling, and builds the characteristic matching point pair of a cloud geometric data;
Step 2, randomly selects a viewpoint as reference view, utilizes the relation homography of the characteristic matching point pair of some cloud geometric data, each viewpoint of computational representation and reference view relative position relation;
Step 3, carries out singular value decomposition to relation homography, asks for relation translation vector and the relation rotating vector of characteristic matching point pair between each viewpoint and reference view;
Step 4, according to relation translation vector and relation rotating vector, is indicated the some cloud geometric data in each viewpoint under reference view coordinate system, completes an automatic Mosaic for cloud geometric data.
2. the automatic Mosaic algorithm of the some cloud geometric data based on multi-view image three-dimensional modeling as claimed in claim 1, it is characterised in that the number of different points of view is at least 6.
3. the automatic Mosaic algorithm of the some cloud geometric data based on multi-view image three-dimensional modeling as claimed in claim 2, it is characterised in that the number of characteristic matching point pair is 100～120.
4. the automatic Mosaic algorithm of the some cloud geometric data based on multi-view image three-dimensional modeling as claimed in claim 3, it is characterised in that the number of characteristic matching point pair is 100.
5. the automatic Mosaic algorithm of the some cloud geometric data based on multi-view image three-dimensional modeling as claimed in claim 4, it is characterised in that use optimization mechanism in described step 2 during calculated relationship homography.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610015916.3A CN105678833A (en) | 2016-01-11 | 2016-01-11 | Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610015916.3A CN105678833A (en) | 2016-01-11 | 2016-01-11 | Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105678833A true CN105678833A (en) | 2016-06-15 |
Family
ID=56299982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610015916.3A Pending CN105678833A (en) | 2016-01-11 | 2016-01-11 | Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105678833A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019041794A1 (en) * | 2017-08-30 | 2019-03-07 | 深圳中科飞测科技有限公司 | Distortion correction method and apparatus for three-dimensional measurement, and terminal device and storage medium |
| CN109974707A (en) * | 2019-03-19 | 2019-07-05 | 重庆邮电大学 | A kind of indoor mobile robot vision navigation method based on improvement cloud matching algorithm |
| CN111536871A (en) * | 2020-05-07 | 2020-08-14 | 武汉大势智慧科技有限公司 | Accurate calculation method for volume variation of multi-temporal photogrammetric data |
| CN111540040A (en) * | 2020-04-20 | 2020-08-14 | 上海曼恒数字技术股份有限公司 | Point cloud data-based model construction method and device and storage medium |
| CN112861674A (en) * | 2021-01-28 | 2021-05-28 | 中振同辂(江苏)机器人有限公司 | Point cloud optimization method based on ground features and computer readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103075977A (en) * | 2012-12-28 | 2013-05-01 | 浙江大学 | Automatic combining algorithm for point cloud data in binocular stereoscopic vision system |
| CN104809759A (en) * | 2015-04-03 | 2015-07-29 | 哈尔滨工业大学深圳研究生院 | Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter |
| US20150317821A1 (en) * | 2014-04-30 | 2015-11-05 | Seiko Epson Corporation | Geodesic Distance Based Primitive Segmentation and Fitting for 3D Modeling of Non-Rigid Objects from 2D Images |
-
2016
- 2016-01-11 CN CN201610015916.3A patent/CN105678833A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103075977A (en) * | 2012-12-28 | 2013-05-01 | 浙江大学 | Automatic combining algorithm for point cloud data in binocular stereoscopic vision system |
| US20150317821A1 (en) * | 2014-04-30 | 2015-11-05 | Seiko Epson Corporation | Geodesic Distance Based Primitive Segmentation and Fitting for 3D Modeling of Non-Rigid Objects from 2D Images |
| CN104809759A (en) * | 2015-04-03 | 2015-07-29 | 哈尔滨工业大学深圳研究生院 | Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019041794A1 (en) * | 2017-08-30 | 2019-03-07 | 深圳中科飞测科技有限公司 | Distortion correction method and apparatus for three-dimensional measurement, and terminal device and storage medium |
| CN109974707A (en) * | 2019-03-19 | 2019-07-05 | 重庆邮电大学 | A kind of indoor mobile robot vision navigation method based on improvement cloud matching algorithm |
| CN111540040A (en) * | 2020-04-20 | 2020-08-14 | 上海曼恒数字技术股份有限公司 | Point cloud data-based model construction method and device and storage medium |
| CN111536871A (en) * | 2020-05-07 | 2020-08-14 | 武汉大势智慧科技有限公司 | Accurate calculation method for volume variation of multi-temporal photogrammetric data |
| CN112861674A (en) * | 2021-01-28 | 2021-05-28 | 中振同辂(江苏)机器人有限公司 | Point cloud optimization method based on ground features and computer readable storage medium |
| CN112861674B (en) * | 2021-01-28 | 2024-09-06 | 中振同辂(江苏)机器人有限公司 | Point cloud optimization method based on ground characteristics and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104596502B (en) | Object posture measuring method based on CAD model and monocular vision | |
| CN107767442A (en) | A kind of foot type three-dimensional reconstruction and measuring method based on Kinect and binocular vision | |
| CN104299261B (en) | Three-dimensional imaging method and system for human body | |
| CN105678833A (en) | Point cloud geometrical data automatic splicing algorithm based on multi-view image three-dimensional modeling | |
| CN103604417B (en) | The multi-view images bi-directional matching strategy that object space is information constrained | |
| CN103252778A (en) | Apparatus for estimating the robot pose and method thereof | |
| CN103075977B (en) | The automatic splicing method of the cloud data in Binocular Stereo Vision System | |
| CN106340059B (en) | Automatic splicing method based on three-dimensional modeling of multi-body feeling acquisition equipment | |
| CN102778224B (en) | Method for aerophotogrammetric bundle adjustment based on parameterization of polar coordinates | |
| CN106295512A (en) | Many correction line indoor vision data base construction method based on mark and indoor orientation method | |
| US6175648B1 (en) | Process for producing cartographic data by stereo vision | |
| CN101149836A (en) | Three-dimensional reconfiguration double pick-up camera calibration method | |
| CN101901502A (en) | Global optimal registration method of multi-viewpoint cloud data during optical three-dimensional measurement | |
| CA2757313C (en) | Stereoscopic measurement system and method | |
| CN104504691A (en) | Camera position and posture measuring method on basis of low-rank textures | |
| CN113487726B (en) | Motion capture system and method | |
| CN104574273A (en) | Point cloud registration system and method | |
| Jiang et al. | Learned local features for structure from motion of uav images: A comparative evaluation | |
| CN108022308A (en) | A kind of facial alignment schemes based on three-dimensional face model fitting | |
| Wang et al. | Single view based measurement on space planes | |
| Sun et al. | A new camera calibration method for multi-camera localization | |
| EP2283314B1 (en) | Stereoscopic measurement system and method | |
| CN114511637A (en) | Weak-feature object image three-dimensional reconstruction system and method based on strong feature construction | |
| Zhang et al. | Point cloud registration with 2D and 3D fusion information on mobile robot integrated vision system | |
| Liu et al. | A 3D spherical panorama modeling method based on double projective geometry |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160615 |
|
| RJ01 | Rejection of invention patent application after publication |