CN103729882A - Point cloud relative pose estimation method based on three-dimensional curve matching - Google Patents
Point cloud relative pose estimation method based on three-dimensional curve matching Download PDFInfo
- Publication number
- CN103729882A CN103729882A CN201310742491.2A CN201310742491A CN103729882A CN 103729882 A CN103729882 A CN 103729882A CN 201310742491 A CN201310742491 A CN 201310742491A CN 103729882 A CN103729882 A CN 103729882A
- Authority
- CN
- China
- Prior art keywords
- point
- curve
- dimensional
- point set
- 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.)
- Granted
Links
Images
Abstract
The invention discloses a point cloud relative pose estimation method based on three-dimensional curve matching. The method includes the steps that the edge contours of two frames of input point clouds are extracted, and three-dimensional point sets (img file='dest_path_image 002. TIF' wi='13' he='16'/) and (img file='dest_path_image 004. TIF' wi='14' he='16'/) on behalf of the three-dimensional edge contour curve are obtained; according to the three-dimensional point sets (img file='915081 dest_path_image 002. TIF' wi='14' he='17'/) and (img file='301063 dest_path_image 004. TIF' wi='14' he='16'/), orderly point sets (img file='dest_path_image 006. TIF' wi='26' he='20'/) are obtained through extraction; each orderly point set (img file='dest_path_image 008. TIF' wi='14' he='20'/) or (img file='dest_path_image 010. TIF' wi='15' he='21'/) is decomposed, each subset is on behalf of one arc section after decomposition, and a feature description vector of each arc section is established; the feature description vectors of the edge curve arc sections of the two frames of point clouds are matched to establish a matching relation, and the relative pose transformational matrixes (img file='dest_path_image 012. TIF' wi='39' he='18'/) of the two frames of point clouds are calculated. The point cloud relative pose estimation method based on three-dimensional curve matching does not depend on an initial pose, only an environmental contour is used for matching, and the relative pose is estimated; only the edge point sets are processed, and calculation is lowered; the method can be suitable for chaff interferent in which point clouds exist.
Description
Technical field
The present invention relates to three-dimensional environment reconstruction field, relate in particular to a kind of some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling.
Background technology
Existing some cloud Relative attitude and displacement estimation method is mainly divided into two classes.One class is based on a cloud characteristic matching, and shortcoming is that the impact of receptor site cloud variable density is very large, and the density variation probability that can directly lead to errors improves.The another kind of direct coupling based on a cloud, shortcoming is to rely on initial pose, and operand is large.
For the coupling of three-dimensional curve, existing method, also mainly based on convolution, is difficult to be applicable to the coupling of many curves to many curves.
Summary of the invention
The object of the invention is to overcome the deficiencies in the prior art, a kind of some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling is provided.
Point cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling is:
1) extract respectively the edge contour of two frame input point clouds, obtain representing the three-dimensional point set of three-dimensional edges contour curve
,
;
2) respectively for three-dimensional point set
,
extraction obtains the set of orderly point set
,
, each orderly point set
or
all represented a three-dimensional curve;
3) by each orderly point set
or
decompose, each subset after decomposition represents an arc section, sets up the feature description vectors of every section of circular arc;
4) the feature description vectors of coupling two frame point cloud boundary curve arc sections, sets up matching relationship, calculates the relative pose transformation matrix of two frame point clouds
,
for rotation matrix,
for translation matrix.
Described step 1) is: a) for unordered some cloud, search for each point around radius be less than
quantity in scope is no more than
all nearest neighbor points, obtain point set
, be point set
fit Plane
, with point set
in plane
interior subpoint position is independent variable, point set
to plane
distance be functional value, matching Binary quadratic functions
, obtain Hessian matrix
, calculate Hessian matrix
eigenwert
with
, suppose
if,
and
, wherein
,
be respectively respective threshold, think that this point is marginal point; B), for orderly some cloud, depth map, utilizes Canny algorithm to extract edge point set.
Described is point set
fit Plane
method be: calculate point set
average, obtain plane
center
; Calculate
proper vector, its minimal eigenvalue characteristic of correspondence vector is plane
normal vector
; Plane
center
and normal vector
represented that one through center
, normal vector is
plane.
Described matching Binary quadratic functions
and obtain Hessian matrix
method be: for point set
middle every bit
, suppose
two maximum eigenwert characteristic of correspondence vectors are respectively
with
, calculate one with
,
for independent variable,
for the key-value pair of value,
One group of final formation
to
key assignments mapping, and utilize least square method to ask for Hessian matrix
Described step 2) be: use region growing algorithm, with at every turn from three-dimensional point set
or
in the region growing Seed Points chosen be at random initial, the represented curve single order of the point set of take is condition continuously, by constantly receiving the new point that meets condition of growth, expand in an orderly manner curve point set, if can not expand again, separately get Seed Points and expand another curve point set, until institute is a little complete by expansion.
The described curve single order condition of continuity is:
Wherein
for two Neighbor Points in a cloud
,
between distance,
,
be respectively at 2 from the distance of true origin,
for threshold value.
Described step 3) is: for each, represent the orderly point set of three-dimensional curve
or
, calculate successively each point
locate bent curvature of a curve
with turn round rate
,
Wherein
for point is concentrated in order
next point, subsequently the orderly point set that represents three-dimensional curve is split into some subsets, the point in each subset has identical curvature of curve, and the rate of turning round is zero, this subset has represented a circular arc, remembers that this circular arc curvature is
, arc length is
, centre coordinate is
, according to right-handed coordinate system rule, take arc method direction, circular arc starting point cutting method to and the direction vertical with this both direction as three axles, set up coordinate system, remember that this coordinate system and world coordinate system angulation vector are
, define the feature description vectors of this circular arc
for
Described step 4) is: for the feature description vectors of two frame point cloud boundary curve arc sections, set up k-d tree, in k-d tree, set up the matching relationship between two frame point cloud boundary curve arc sections, use random sampling consistency algorithm, remove part erroneous matching, set up cost function
Wherein
be one group of matching relationship in the matching relationship between two frame point cloud boundary curve arc sections,
for the matching relationship of all existence,
,
be respectively the feature description vectors of one section of arc section in two frame point clouds,
it is assessment feature description vectors
,
between the function of distance, with the relative pose transition matrix of two frame point clouds
it is relevant,
be defined as
Finally, Optimization Solution
The present invention compared with prior art, the beneficial effect having:
1. do not rely on initial bit appearance, only the profile by environment mates, and estimates relative pose;
2. only process edge point set, reduced operand;
3. adapt to the coupling of many curves to many curves, can resist the impact that in a cloud, chaff interference causes.
Accompanying drawing explanation
Fig. 1 is the some cloud Relative attitude and displacement estimation method operation steps schematic diagram based on three-dimensional curve coupling;
Fig. 2 is the some cloud Relative attitude and displacement estimation method implementation result figure based on three-dimensional curve coupling;
Fig. 3 is the some cloud Relative attitude and displacement estimation method application example figure based on three-dimensional curve coupling.
Embodiment
Point cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling is by being some arc sections by the edge contour curve separating of three-dimensional point cloud, definition be take arc section as basic three-dimensional curve vector quantization description, for two frame cloud datas, comparing and mate the vector quantization of its edge contour curve describes, the relative pose of asking for two frame cloud datas, concrete steps are as follows:
1) extract respectively the edge contour of two frame input point clouds, obtain representing the three-dimensional point set of three-dimensional edges contour curve
,
;
2) respectively for three-dimensional point set
,
extraction obtains the set of orderly point set
,
, each orderly point set
or
all represented a three-dimensional curve;
3) by each orderly point set
or
decompose, each subset after decomposition represents an arc section, sets up the feature description vectors of every section of circular arc;
4) the feature description vectors of coupling two frame point cloud boundary curve arc sections, sets up matching relationship, calculates the relative pose transformation matrix of two frame point clouds
,
for rotation matrix,
for translation matrix.
Described step 1) is: a) for unordered some cloud, search for each point around radius be less than
quantity in scope is no more than
all nearest neighbor points, obtain point set
, be point set
fit Plane
, with point set
in plane
interior subpoint position is independent variable, point set
to plane
distance be functional value, matching Binary quadratic functions
, obtain Hessian matrix
, calculate Hessian matrix
eigenwert
with
, suppose
if,
and
, wherein
,
be respectively respective threshold, think that this point is marginal point; B) for orderly some cloud, it is depth map, utilize Canny algorithm (Canny J. A computational approach to edge detection[J]. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 1986 (6): 679-698.) extract edge point set.
Described is point set
fit Plane
method be: calculate point set
average, obtain plane
center
; Calculate
proper vector, its minimal eigenvalue characteristic of correspondence vector is plane
normal vector
; Plane
center
and normal vector
represented that one through center
, normal vector is
plane.
Described matching Binary quadratic functions
and obtain Hessian matrix
method be: for point set
middle every bit
, suppose
two maximum eigenwert characteristic of correspondence vectors are respectively
with
, calculate one with
,
for independent variable,
for the key-value pair of value,
One group of final formation
to
key assignments mapping, and utilize least square method to ask for Hessian matrix
,
Described step 2) be: use region growing algorithm (Adams R, Bischof L. Seeded region growing[J]. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 1994,16 (6): 641-647.), with at every turn from three-dimensional point set
or
in the region growing Seed Points chosen be at random initial, the represented curve single order of the point set of take is condition continuously, by constantly receiving the new point that meets condition of growth, expand in an orderly manner curve point set, if can not expand again, separately get Seed Points and expand another curve point set, until institute is a little complete by expansion.
The described curve single order condition of continuity is:
Wherein
for two Neighbor Points in a cloud
,
between distance,
,
be respectively at 2 from the distance of true origin,
for threshold value.
Described step 3) is: for each, represent the orderly point set of three-dimensional curve
or
, calculate successively each point
locate bent curvature of a curve
with turn round rate
,
Wherein
for point is concentrated in order
next point, subsequently the orderly point set that represents three-dimensional curve is split into some subsets, the point in each subset has identical curvature of curve, and the rate of turning round is zero, this subset has represented a circular arc, remembers that this circular arc curvature is
, arc length is
, centre coordinate is
, according to right-handed coordinate system rule, take arc method direction, circular arc starting point cutting method to and the direction vertical with this both direction as three axles, set up coordinate system, remember that this coordinate system and world coordinate system angulation vector are
, define the feature description vectors of this circular arc
for
Described step 4) is: for the feature description vectors of two frame point cloud boundary curve arc sections, set up k-d tree, in k-d tree, set up the matching relationship between two frame point cloud boundary curve arc sections, use random sampling consistency algorithm (Fischler M A, Bolles R C. Random sample consensus:a paradigm for model fitting with applications to image analysis and automated cartography[J]. Communications of the ACM, 1981, 24 (6): 381-395.), remove part erroneous matching, set up cost function
Wherein
be one group of matching relationship in the matching relationship between two frame point cloud boundary curve arc sections,
for the matching relationship of all existence,
,
be respectively the feature description vectors of one section of arc section in two frame point clouds,
it is assessment feature description vectors
,
between the function of distance, with the relative pose transition matrix of two frame point clouds
it is relevant,
be defined as
Finally, Optimization Solution
Claims (8)
1. a some cloud Relative attitude and displacement estimation method of mating based on three-dimensional curve, is characterized in that its step is as follows:
1) extract respectively the edge contour of two frame input point clouds, obtain representing the three-dimensional point set of three-dimensional edges contour curve
,
;
2) respectively for three-dimensional point set
,
extraction obtains the set of orderly point set
,
, each orderly point set
or
all represented a three-dimensional curve;
3) by each orderly point set
or
decompose, each subset after decomposition represents an arc section, sets up the feature description vectors of every section of circular arc;
2. the some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling according to claim 1, is characterized in that, described step 1) is: a) for unordered some cloud, search for each point around radius be less than
quantity in scope is no more than
all nearest neighbor points, obtain point set
, be point set
fit Plane
, with point set
in plane
interior subpoint position is independent variable, point set
to plane
distance be functional value, matching Binary quadratic functions
, obtain Hessian matrix
, calculate Hessian matrix
eigenwert
with
, suppose
if,
and
, wherein
,
be respectively respective threshold, think that this point is marginal point; B), for orderly some cloud, depth map, utilizes Canny algorithm to extract edge point set.
3. the some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling according to claim 2, is characterized in that, described is point set
fit Plane
method be: calculate point set
average, obtain plane
center
; Calculate
proper vector, its minimal eigenvalue characteristic of correspondence vector is plane
normal vector
; Plane
center
and normal vector
represented that one through center
, normal vector is
plane.
4. the some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling according to claim 2, is characterized in that described matching Binary quadratic functions
and obtain Hessian matrix
method be: for point set
middle every bit
, suppose
two maximum eigenwert characteristic of correspondence vectors are respectively
with
, calculate one with
,
for independent variable,
for the key-value pair of value,
One group of final formation
to
key assignments mapping, and utilize least square method to ask for Hessian matrix
5. the some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling according to claim 1, is characterized in that described step 2) be: region growing algorithm used, with at every turn from three-dimensional point set
or
in the region growing Seed Points chosen be at random initial, the represented curve single order of the point set of take is condition continuously, by constantly receiving the new point that meets condition of growth, expand in an orderly manner curve point set, if can not expand again, separately get Seed Points and expand another curve point set, until institute is a little complete by expansion.
6. the some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling according to claim 5, is characterized in that, the described curve single order condition of continuity is:
7. the some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling according to claim 1, is characterized in that, described step 3) is: for each, represent the orderly point set of three-dimensional curve
or
, calculate successively each point
locate bent curvature of a curve
with turn round rate
,
Wherein
for point is concentrated in order
next point, subsequently the orderly point set that represents three-dimensional curve is split into some subsets, the point in each subset has identical curvature of curve, and the rate of turning round is zero, this subset has represented a circular arc, remembers that this circular arc curvature is
, arc length is
, centre coordinate is
, according to right-handed coordinate system rule, take arc method direction, circular arc starting point cutting method to and the direction vertical with this both direction as three axles, set up coordinate system, remember that this coordinate system and world coordinate system angulation vector are
, define the feature description vectors of this circular arc
for
8. the some cloud Relative attitude and displacement estimation method based on three-dimensional curve coupling according to claim 1, it is characterized in that, described step 4) is: for the feature description vectors of two frame point cloud boundary curve arc sections, set up k-d tree, in k-d tree, set up the matching relationship between two frame point cloud boundary curve arc sections, use random sampling consistency algorithm, remove part erroneous matching, set up cost function
Wherein
be one group of matching relationship in the matching relationship between two frame point cloud boundary curve arc sections,
for the matching relationship of all existence,
,
be respectively the feature description vectors of one section of arc section in two frame point clouds,
it is assessment feature description vectors
,
between the function of distance, with the relative pose transition matrix of two frame point clouds
it is relevant,
be defined as
Finally, Optimization Solution
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310742491.2A CN103729882B (en) | 2013-12-30 | 2013-12-30 | A kind of some cloud relative pose estimation method based on three-dimensional curve coupling |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310742491.2A CN103729882B (en) | 2013-12-30 | 2013-12-30 | A kind of some cloud relative pose estimation method based on three-dimensional curve coupling |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103729882A true CN103729882A (en) | 2014-04-16 |
CN103729882B CN103729882B (en) | 2016-09-28 |
Family
ID=50453943
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310742491.2A Active CN103729882B (en) | 2013-12-30 | 2013-12-30 | A kind of some cloud relative pose estimation method based on three-dimensional curve coupling |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103729882B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104123546A (en) * | 2014-07-25 | 2014-10-29 | 黑龙江省科学院自动化研究所 | Multi-dimensional feature extraction based bullet trace comparison method |
CN105021124A (en) * | 2015-04-16 | 2015-11-04 | 华南农业大学 | Planar component three-dimensional position and normal vector calculation method based on depth map |
CN105509748A (en) * | 2015-12-29 | 2016-04-20 | 深圳先进技术研究院 | Navigation method and apparatus for robot |
CN109613543A (en) * | 2018-12-06 | 2019-04-12 | 深圳前海达闼云端智能科技有限公司 | Method and device for correcting laser point cloud data, storage medium and electronic equipment |
CN110186467A (en) * | 2018-02-23 | 2019-08-30 | 通用汽车环球科技运作有限责任公司 | Group's sensing points cloud map |
CN113763560A (en) * | 2021-08-02 | 2021-12-07 | 纵目科技(上海)股份有限公司 | Method, system and equipment for generating point cloud data and computer readable storage medium |
CN114943058A (en) * | 2022-04-18 | 2022-08-26 | 哈尔滨理工大学 | Orthodontic arch wire error fluctuation degree evaluation method based on position error judgment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130009954A1 (en) * | 2010-05-20 | 2013-01-10 | Microsoft Corporation | Spatially registering user photographs |
CN103150747A (en) * | 2012-11-16 | 2013-06-12 | 佳都新太科技股份有限公司 | Point cloud registration algorithm based on topological characteristic |
CN103383776A (en) * | 2013-07-14 | 2013-11-06 | 浙江大学 | Progressive type three-dimensional matching algorithm based on sectional matching and bayes estimation |
-
2013
- 2013-12-30 CN CN201310742491.2A patent/CN103729882B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130009954A1 (en) * | 2010-05-20 | 2013-01-10 | Microsoft Corporation | Spatially registering user photographs |
CN103150747A (en) * | 2012-11-16 | 2013-06-12 | 佳都新太科技股份有限公司 | Point cloud registration algorithm based on topological characteristic |
CN103383776A (en) * | 2013-07-14 | 2013-11-06 | 浙江大学 | Progressive type three-dimensional matching algorithm based on sectional matching and bayes estimation |
Non-Patent Citations (4)
Title |
---|
刘洁 等: "基于空问曲线重建的点云模型精化方法", 《测绘通报》 * |
吴禄慎: "基于特征点的改进ICP三维点云配准技术", 《南昌大学学报·工科版》 * |
牛丹华 等: "改进的多视图点云数据配准方法", 《工具技术》 * |
葛宝臻: "基于曲率图的三维点云数据配准", 《天津大学学报(自然科学与工程技术版)》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104123546A (en) * | 2014-07-25 | 2014-10-29 | 黑龙江省科学院自动化研究所 | Multi-dimensional feature extraction based bullet trace comparison method |
CN105021124A (en) * | 2015-04-16 | 2015-11-04 | 华南农业大学 | Planar component three-dimensional position and normal vector calculation method based on depth map |
CN105021124B (en) * | 2015-04-16 | 2018-02-16 | 华南农业大学 | A kind of planar part three-dimensional position and normal vector computational methods based on depth map |
CN105509748A (en) * | 2015-12-29 | 2016-04-20 | 深圳先进技术研究院 | Navigation method and apparatus for robot |
CN110186467A (en) * | 2018-02-23 | 2019-08-30 | 通用汽车环球科技运作有限责任公司 | Group's sensing points cloud map |
CN109613543A (en) * | 2018-12-06 | 2019-04-12 | 深圳前海达闼云端智能科技有限公司 | Method and device for correcting laser point cloud data, storage medium and electronic equipment |
CN113763560A (en) * | 2021-08-02 | 2021-12-07 | 纵目科技(上海)股份有限公司 | Method, system and equipment for generating point cloud data and computer readable storage medium |
CN113763560B (en) * | 2021-08-02 | 2024-02-09 | 纵目科技(上海)股份有限公司 | Method, system, equipment and computer readable storage medium for generating point cloud data |
CN114943058A (en) * | 2022-04-18 | 2022-08-26 | 哈尔滨理工大学 | Orthodontic arch wire error fluctuation degree evaluation method based on position error judgment |
Also Published As
Publication number | Publication date |
---|---|
CN103729882B (en) | 2016-09-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103729882A (en) | Point cloud relative pose estimation method based on three-dimensional curve matching | |
CN110443836B (en) | Point cloud data automatic registration method and device based on plane features | |
CN111508074B (en) | Three-dimensional building model simplification method based on roof contour line | |
CN103727930B (en) | A kind of laser range finder based on edge matching and camera relative pose scaling method | |
CN108010081B (en) | RGB-D visual odometer method based on Census transformation and local graph optimization | |
CN109685080B (en) | Multi-scale plane extraction method based on Hough transformation and region growth | |
TWI497450B (en) | Visual object tracking method | |
CN107818598B (en) | Three-dimensional point cloud map fusion method based on visual correction | |
CN105261047A (en) | Docking ring circle center extraction method based on close-range short-arc image | |
CN109146001B (en) | Multi-view ISAR image fusion method | |
CN105160686B (en) | A kind of low latitude various visual angles Remote Sensing Images Matching Method based on improvement SIFT operators | |
CN108932734B (en) | Monocular image depth recovery method and device and computer equipment | |
CN111998862B (en) | BNN-based dense binocular SLAM method | |
CN104050681A (en) | Road vanishing point detection method based on video images | |
CN103279955B (en) | Image matching method and system | |
CN111709450B (en) | Point cloud normal vector estimation method and system based on multi-scale feature fusion | |
CN104657986A (en) | Quasi-dense matching extension method based on subspace fusion and consistency constraint | |
CN104881029A (en) | Mobile robot navigation method based on one point RANSAC and FAST algorithm | |
CN104463953A (en) | Three-dimensional reconstruction method based on inertial measurement unit and RGB-D sensor | |
CN108805915A (en) | A kind of close-range image provincial characteristics matching process of anti-visual angle change | |
CN105469110A (en) | Non-rigid transformation image characteristic matching method based on local linear transfer and system | |
CN114202632A (en) | Grid linear structure recovery method and device, electronic equipment and storage medium | |
CN110967020B (en) | Simultaneous drawing and positioning method for port automatic driving | |
Zhu et al. | An improved RANSAC algorithm based on similar structure constraints | |
CN111664845B (en) | Traffic sign positioning and visual map making method and device and positioning system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20180806 Address after: 310052 208, room 6, 1197 Binan Road, Binjiang District, Hangzhou, Zhejiang. Patentee after: Which Hangzhou science and Technology Co Ltd Address before: 310027 No. 38, Zhejiang Road, Hangzhou, Zhejiang, Xihu District Patentee before: Zhejiang University |