CN103456016A - Body-feeling camera network calibration method unrelated to visual angles - Google Patents

Body-feeling camera network calibration method unrelated to visual angles Download PDF

Info

Publication number
CN103456016A
CN103456016A CN2013104039881A CN201310403988A CN103456016A CN 103456016 A CN103456016 A CN 103456016A CN 2013104039881 A CN2013104039881 A CN 2013104039881A CN 201310403988 A CN201310403988 A CN 201310403988A CN 103456016 A CN103456016 A CN 103456016A
Authority
CN
China
Prior art keywords
coordinate system
video camera
human skeleton
rotation matrix
frame
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
Application number
CN2013104039881A
Other languages
Chinese (zh)
Other versions
CN103456016B (en
Inventor
陈启军
韩云
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
Original Assignee
Tongji University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tongji University filed Critical Tongji University
Priority to CN201310403988.1A priority Critical patent/CN103456016B/en
Publication of CN103456016A publication Critical patent/CN103456016A/en
Application granted granted Critical
Publication of CN103456016B publication Critical patent/CN103456016B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Image Analysis (AREA)
  • Image Processing (AREA)
  • Closed-Circuit Television Systems (AREA)
  • Length Measuring Devices By Optical Means (AREA)

Abstract

The invention relates to a body-feeling camera network calibration method unrelated to visual angles. The method comprises the steps that 1) the coordinate system of one of a plurality of cameras which are installed is selected to serve as a reference coordinate system, and other coordinate systems are non-reference coordinate systems; 2) the cameras obtain and display skeleton position information of the same person simultaneously; 3) a human skeleton coordinate system is built, and the cameras calculate rotation matrixes and horizontally-moving vectors of the cameras relative to the human skeleton coordinate system respectively according to the skeleton position information obtained in a detected mode; 4) according to the rotation matrixes and the horizontally-moving vectors obtained in the step 3), rotation matrixes and horizontally-moving vectors of all the non-reference coordinate systems relative to the reference coordinate system are calculated; 5) according to the rotation matrixes and the horizontally-moving vectors obtained in the step 4), the skeleton position information detected the cameras in the non-reference coordinate systems is mapped in the reference coordinate system. Compared with the prior art, the body-feeling camera network calibration method has the advantages that operation is simple, practicality is strong, and other auxiliary markers do not need to be prepared specially.

Description

A kind of body sense video camera network mark of view angle-independent is determined method
Technical field
The present invention relates to a kind of camera marking method, the body sense video camera network mark that especially relates to a kind of view angle-independent is determined method.
Background technology
The demarcation of camera can be divided into two classes: a class is the demarcation of one camera; Another kind of is the global calibration of polyphaser.The demarcation of one camera refers to the Relation Parameters solved between every camera inside and outside parameter and camera.Method commonly used is the chessboard calibration method of Zhengyou Zhang proposition and two standardizations of RAC that Tsai proposes.Wherein, the chessboard calibration method that Zhang proposes, serve as a mark a little by the summit by artificial mark chessboard black and white chess lattice, and stated accuracy is high, has become a kind of method of main flow.The global calibration of polyphaser refers to the data unification of all cameras in the overall frame of reference, obtains the relative coordinate between camera, namely translation vector T and the rotation matrix R between camera coordinates system.Main thought is to find out the 3D gauge point jointly detected between camera to calculate the spatial relationship between camera.The point cloud that adopts kinect to detect at present is gauge point, by the ICP method, calculates.The major defect of said method is that calculated amount is large, needs extra auxiliary reference thing.
Summary of the invention
Purpose of the present invention is exactly in order to overcome the defect that above-mentioned prior art exists, to provide a kind of body sense video camera network mark of simple to operate, real-time view angle-independent to determine method.
Purpose of the present invention can be achieved through the following technical solutions:
A kind of body sense video camera network mark of view angle-independent is determined method, and the method comprises the following steps:
1) coordinate at the selected video camera place of the multiple cameras of installing is the frame of reference, and all the other are the non-frame of reference;
2) multiple cameras obtains the frame position information of same person simultaneously and shows;
3) set up the human skeleton coordinate system, multiple cameras respectively according to detect frame position information calculate rotation matrix and the translation vector of the relative human skeleton coordinate system of each video camera;
4) according to step 3) rotation matrix and the translation vector that obtain, calculate rotation matrix and the translation vector of each non-frame of reference relative datum coordinate system;
5) according to step 4) rotation matrix and the translation vector that obtain, the frame position information of non-frame of reference place video camera detecting is mapped in the frame of reference, and shows, complete demarcation.
Described step 3), in, the positional information of the left shoulder of the human body of take, right shoulder and barycenter is fundamental construction human skeleton coordinate system, is specially:
The left shoulder of take is X-direction to the direction of right shoulder, and the barycenter of take is Y direction to the central point of right and left shoulders, by the right-hand rule, determines Z-direction.
Described step 3), in, the peaceful amount of shifting to of rotation matrix of calculating the relative human skeleton coordinate system of each video camera is specially:
Obtain the left shoulder of skeleton, right shoulder, barycenter vector and the vector in the human skeleton coordinate system thereof in the coordinate system of video camera S place, calculate the rotation matrix of the relative human skeleton coordinate system of video camera S according to following formula
Figure BDA0000378277360000021
and translation vector
Figure BDA0000378277360000022
Figure BDA0000378277360000023
sp is the expression vector of some P correspondence in the coordinate system of video camera S place, the peoplethe expression vector that P is same point P correspondence in the human skeleton coordinate system,
Figure BDA0000378277360000024
the rotation matrix that means the relative video camera S of human skeleton coordinate system place coordinate system,
Figure BDA0000378277360000025
the translation vector that means the relative video camera S of human skeleton coordinate system place coordinate system.
Described step 4), in, the peaceful amount of shifting to of rotation matrix of calculating each non-frame of reference relative datum coordinate system is specially:
The expression vector of some P correspondence in the human skeleton coordinate system the peoplep is for being expressed as:
basep is the expression vector of some P correspondence in the coordinate system of fiducial cameras place,
Figure BDA0000378277360000027
the rotation matrix that means the relative human skeleton coordinate system of fiducial cameras,
Figure BDA0000378277360000028
mean the translation vector of fiducial cameras with respect to the human skeleton coordinate system, non-p is the expression vector of some P correspondence in the coordinate system of non-fiducial cameras place,
Figure BDA0000378277360000029
the rotation matrix that means the relative human skeleton coordinate system of non-fiducial cameras,
Figure BDA00003782773600000210
mean the translation vector of non-fiducial cameras with respect to the human skeleton coordinate system,
The rotation matrix of non-frame of reference relative datum coordinate system
Figure BDA00003782773600000212
the translation vector of non-frame of reference relative datum coordinate system
Figure BDA0000378277360000031
Compared with prior art, it is that reference point builds the coordinate system for the view angle-independent of human skeleton that the present invention adopts the stable articulation point of human body, aid mark thing simple to operate, real-time, as not need specially to prepare other, and with the view angle-independent of each video camera.
The accompanying drawing explanation
Fig. 1 is schematic flow sheet of the present invention;
Fig. 2 is the human skeleton figure that depth camera detects;
Fig. 3 is for take the human skeleton coordinate system schematic diagram that sane articulation point is fundamental construction.
Embodiment
Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.The present embodiment be take technical solution of the present invention and is implemented as prerequisite, provided detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.
Embodiment
As shown in Figure 1, a kind of body sense video camera network mark of view angle-independent is determined method, and the sane nominal volume sense video camera net fast of take is target, and sane skeleton joint point is reference point to propose to take human body, build the human skeleton coordinate system of view angle-independent, and adopt the method based on the coordinate system conversion to realize.The method comprises the following steps:
In step S101, the coordinate at the selected video camera place of the multiple cameras of installing is the frame of reference, and all the other are the non-frame of reference, and the present embodiment be take 2 as example, is assumed to A and B, and the coordinate at selected video camera A place is the frame of reference;
In step S102,2 video cameras obtain the frame position information of same person simultaneously and show;
In step S103, set up the human skeleton coordinate system, 2 video cameras respectively according to detect frame position information calculate the rotation matrix of the relative human skeleton coordinate system of each video camera
Figure BDA0000378277360000033
and translation vector
Figure BDA0000378277360000034
Figure BDA0000378277360000035
In step S104, according to rotation matrix and the translation vector of step S103 acquisition, calculate rotation matrix and the translation vector of each non-frame of reference relative datum coordinate system;
In step S105, the rotation matrix and the translation vector that according to step S104, obtain, the frame position information of non-frame of reference place video camera (referring to B here) detecting is mapped in the frame of reference (referring to A here), and shows, complete demarcation.
Take the view angle-independent switch technology that shoulder bone and barycenter be benchmark:
Depth camera kinect, under the assistance of software OpenNI, can effectively obtain 3D coordinate and the direction of 15 articulation points of human body, and articulation point is shown in Fig. 2.By long observation and experiment, find, in 15 articulation points obtaining, stable articulation point mainly contains 4, is respectively left shoulder, right shoulder, head and barycenter.Simultaneously, the upper part of the body torso portion of also finding human body is all very stable, is difficult to change, and the result that kinect detects is also the same.So the present invention proposes to take left shoulder, right shoulder and barycenter as fundamental construction human skeleton coordinate system.The left shoulder of take is X-direction to the direction of right shoulder, and the barycenter of take is Y-direction to the central point of right and left shoulders, by the right-hand rule, determines the Z direction, specifically sees Fig. 3.
To calculate the rotation matrix of the relative human skeleton coordinate system of video camera A
Figure BDA0000378277360000041
and translation vector
Figure BDA0000378277360000042
for example,
Figure BDA0000378277360000043
Wherein, athe expression vector of the correspondence in the coordinate system of camera A place of 1 P in the P representation space, the peoplep means the expression vector of same point P correspondence in the human skeleton coordinate system,
Figure BDA0000378277360000044
the rotation matrix that means the relative camera A of human skeleton coordinate system,
Figure BDA0000378277360000045
mean that the human skeleton coordinate system is with respect to the translation vector at camera A, by measuring, the implication of whole equation expression is that the point in the human skeleton coordinate system is videoed in the coordinate system of camera A place.
The translation vector on formula (1) the right is moved to left, can obtain formula (2):
Suppose:
Figure BDA0000378277360000047
p is known, and, formula (2) is expressed as:
Figure BDA0000378277360000048
Wherein, will
Figure BDA0000378277360000049
be expressed as: if the peoplep=[1 0 0] t:
Figure BDA00003782773600000411
If by right shoulder at the positional representation of human skeleton coordinate system be: (t, 0,0), can obtain:
Figure BDA00003782773600000412
can obtain a b c = P / t · - - - ( 4 )
In like manner, if the peoplep=[0 1 0] t:
Figure BDA00003782773600000413
If by barycenter at the positional representation of human skeleton coordinate system be (0 ,-s, 0):
Figure BDA0000378277360000051
can obtain d e f = P / ( - s ) . - - - ( 5 )
By the right-hand rule, can obtain
Figure BDA0000378277360000053
the 3rd row be just in time the 1st row and the 2nd apposition be listed as:
g h i = b × f - e × c c × d - f × a a × e - d × b ·
:
Figure BDA0000378277360000055
Due to, what need at present is that current coordinate system is mapped to the rotation matrix in the human skeleton coordinate system, what need is (rotation matrix that means the relative human skeleton coordinate system of camera A) can be obtained by the characteristics of rotation matrix:
Figure BDA0000378277360000057
?
Figure BDA0000378277360000058
What in like manner need here to solve is with the human skeleton coordinate, to be the translation vector of benchmark
Figure BDA0000378277360000059
by formula (1), can be obtained
Figure BDA00003782773600000510
and due to
Figure BDA00003782773600000511
therefore:
Figure BDA00003782773600000512
can obtain,
In like manner can obtain the rotation matrix of the relative human skeleton coordinate system of video camera B
Figure BDA00003782773600000514
and translation vector
Figure BDA00003782773600000515
Conversion between coordinate system:
The point that the non-reference coordinate of 2 video cameras is detected is mapped in reference coordinate.In previous step, solved rotation matrix and the translation vector of relative human skeleton coordinate system, the purpose of this step is rotation matrix and the translation vector that solves the relative video camera A of video camera B.Basic ideas are to be that bridge is obtained with the human skeleton coordinate.
Suppose the coordinate system that three coordinate system video camera A are arranged now, the coordinate system of video camera B and the coordinate system of human skeleton.。By video camera A, to people's coordinate system conversion table, be shown:
Figure BDA00003782773600000516
By video camera B, to people's coordinate system conversion table, be shown:
Figure BDA00003782773600000517
Because:
Figure BDA00003782773600000518
so, can obtain
Figure BDA00003782773600000519
further, can obtain:
Figure BDA00003782773600000520
so:
Figure BDA0000378277360000061
Figure BDA0000378277360000062
Finally, will solve
Figure BDA0000378277360000063
with
Figure BDA0000378277360000064
point in non-frame of reference B can be mapped in Basic Reference Coordinate System A, complete the demarcation of camera.

Claims (4)

1. the body sense video camera network mark of a view angle-independent is determined method, it is characterized in that, the method comprises the following steps:
1) coordinate at the selected video camera place of the multiple cameras of installing is the frame of reference, and all the other are the non-frame of reference;
2) multiple cameras obtains the frame position information of same person simultaneously and shows;
3) set up the human skeleton coordinate system, multiple cameras respectively according to detect frame position information calculate rotation matrix and the translation vector of the relative human skeleton coordinate system of each video camera;
4) according to step 3) rotation matrix and the translation vector that obtain, calculate rotation matrix and the translation vector of each non-frame of reference relative datum coordinate system;
5) according to step 4) rotation matrix and the translation vector that obtain, the frame position information of non-frame of reference place video camera detecting is mapped in the frame of reference, and shows, complete demarcation.
2. the body sense video camera network mark of a kind of view angle-independent according to claim 1 is determined method, it is characterized in that described step 3) in, the positional information of the left shoulder of the human body of take, right shoulder and barycenter is fundamental construction human skeleton coordinate system, is specially:
The left shoulder of take is X-direction to the direction of right shoulder, and the barycenter of take is Y direction to the central point of right and left shoulders, by the right-hand rule, determines Z-direction.
3. the body sense video camera network mark of a kind of view angle-independent according to claim 2 is determined method, it is characterized in that described step 3) in, the peaceful amount of shifting to of rotation matrix of calculating the relative human skeleton coordinate system of each video camera is specially:
Obtain the left shoulder of skeleton, right shoulder, barycenter vector and the vector in the human skeleton coordinate system thereof in the coordinate system of video camera S place, calculate the rotation matrix of the relative human skeleton coordinate system of video camera S according to following formula
Figure FDA0000378277350000011
and translation vector
Figure FDA0000378277350000012
Figure FDA0000378277350000013
sp is the expression vector of some P correspondence in the coordinate system of video camera S place, the peoplethe expression vector that P is same point P correspondence in the human skeleton coordinate system,
Figure FDA0000378277350000014
the rotation matrix that means the relative video camera S of human skeleton coordinate system place coordinate system, the translation vector that means the relative video camera S of human skeleton coordinate system place coordinate system.
4. the body sense video camera network mark of a kind of view angle-independent according to claim 3 is determined method, it is characterized in that described step 4) in, the peaceful amount of shifting to of rotation matrix of calculating each non-frame of reference relative datum coordinate system is specially:
The expression vector of some P correspondence in the human skeleton coordinate system the peoplep is for being expressed as:
Figure FDA0000378277350000021
basep is the expression vector of some P correspondence in the coordinate system of fiducial cameras place,
Figure FDA0000378277350000022
the rotation matrix that means the relative human skeleton coordinate system of fiducial cameras, mean the translation vector of fiducial cameras with respect to the human skeleton coordinate system, non-p is the expression vector of some P correspondence in the coordinate system of non-fiducial cameras place,
Figure FDA0000378277350000024
the rotation matrix that means the relative human skeleton coordinate system of non-fiducial cameras,
Figure FDA0000378277350000025
mean the translation vector of non-fiducial cameras with respect to the human skeleton coordinate system,
Figure FDA0000378277350000026
The rotation matrix of non-frame of reference relative datum coordinate system
Figure FDA0000378277350000027
the translation vector of non-frame of reference relative datum coordinate system
Figure FDA0000378277350000028
CN201310403988.1A 2013-09-06 2013-09-06 The body-sensing video camera net scaling method that a kind of visual angle is unrelated Active CN103456016B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310403988.1A CN103456016B (en) 2013-09-06 2013-09-06 The body-sensing video camera net scaling method that a kind of visual angle is unrelated

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310403988.1A CN103456016B (en) 2013-09-06 2013-09-06 The body-sensing video camera net scaling method that a kind of visual angle is unrelated

Publications (2)

Publication Number Publication Date
CN103456016A true CN103456016A (en) 2013-12-18
CN103456016B CN103456016B (en) 2016-07-13

Family

ID=49738344

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310403988.1A Active CN103456016B (en) 2013-09-06 2013-09-06 The body-sensing video camera net scaling method that a kind of visual angle is unrelated

Country Status (1)

Country Link
CN (1) CN103456016B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106060524A (en) * 2016-06-30 2016-10-26 北京邮电大学 Method and device for setting camera
CN109241841A (en) * 2018-08-01 2019-01-18 甘肃未来云数据科技有限公司 The acquisition methods and device of video human movement
CN113077519A (en) * 2021-03-18 2021-07-06 中国电子科技集团公司第五十四研究所 Multi-phase external parameter automatic calibration method based on human skeleton extraction

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101186038A (en) * 2007-12-07 2008-05-28 北京航空航天大学 Method for demarcating robot stretching hand and eye
CN102622766A (en) * 2012-03-01 2012-08-01 西安电子科技大学 Multi-objective optimization multi-lens human motion tracking method
CN102638653A (en) * 2012-03-01 2012-08-15 北京航空航天大学 Automatic face tracing method on basis of Kinect

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101186038A (en) * 2007-12-07 2008-05-28 北京航空航天大学 Method for demarcating robot stretching hand and eye
CN102622766A (en) * 2012-03-01 2012-08-01 西安电子科技大学 Multi-objective optimization multi-lens human motion tracking method
CN102638653A (en) * 2012-03-01 2012-08-15 北京航空航天大学 Automatic face tracing method on basis of Kinect

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
贾倩倩 等: "基于特征点图像序列的多摄像机全局标定", 《清华大学学报(自然科学版)》 *
韩云 等: "以人体骨架为基础的室内实时动作侦测", 《第三十二届中国控制会议论文集》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106060524A (en) * 2016-06-30 2016-10-26 北京邮电大学 Method and device for setting camera
CN106060524B (en) * 2016-06-30 2017-12-29 北京邮电大学 The method to set up and device of a kind of video camera
CN109241841A (en) * 2018-08-01 2019-01-18 甘肃未来云数据科技有限公司 The acquisition methods and device of video human movement
CN109241841B (en) * 2018-08-01 2022-07-05 甘肃未来云数据科技有限公司 Method and device for acquiring video human body actions
CN113077519A (en) * 2021-03-18 2021-07-06 中国电子科技集团公司第五十四研究所 Multi-phase external parameter automatic calibration method based on human skeleton extraction
CN113077519B (en) * 2021-03-18 2022-12-09 中国电子科技集团公司第五十四研究所 Multi-phase external parameter automatic calibration method based on human skeleton extraction

Also Published As

Publication number Publication date
CN103456016B (en) 2016-07-13

Similar Documents

Publication Publication Date Title
CN102155923B (en) Splicing measuring method and system based on three-dimensional target
CN100429476C (en) Double-sensor laser visual measuring system calibrating method
CN104075688B (en) A kind of binocular solid stares the distance-finding method of monitoring system
CN105073348B (en) Robot system and method for calibration
CN109242915A (en) Multicamera system scaling method based on multi-face solid target
CN102980528B (en) Calibration method of pose position-free constraint line laser monocular vision three-dimensional measurement sensor parameters
CN107883870A (en) Overall calibration method based on binocular vision system and laser tracker measuring system
CN104864807B (en) A kind of manipulator hand and eye calibrating method based on active binocular vision
CN107256568B (en) High-precision mechanical arm hand-eye camera calibration method and calibration system
CN103033132B (en) Plane survey method and device based on monocular vision
CN105205824A (en) Multi-camera global calibration method based on high-precision auxiliary cameras and ball targets
CN103353388B (en) A kind of binocular body formula micro imaging system scaling method of tool camera function and device
CN103258329B (en) A kind of camera marking method based on ball one-dimensional
CN1971206A (en) Calibration method for binocular vision sensor based on one-dimension target
CN108269286A (en) Polyphaser pose correlating method based on combination dimensional mark
CN102930548B (en) Utilize the intersecting elliptical linear solution camera intrinsic parameter that two identical
CN104111071B (en) High-precision position posture calculating method based on laser ranging and camera visual fusion
CN101750014A (en) Method for calibrating a camera in an orthogonal three-coordinate measuring machine
CN102359780A (en) Ground target positioning method applied into video monitoring system
CN103679693A (en) Multi-camera single-view calibration device and calibration method thereof
CN105469389A (en) Grid ball target for visual sensor calibration and corresponding calibration method
CN104697463A (en) Blanking feature constraining calibrating method and device for binocular vision sensor
CN102930551B (en) Camera intrinsic parameters determined by utilizing projected coordinate and epipolar line of centres of circles
CN104655106B (en) Autonomous positioning based on GPS RTK and full-view image orients plotting method
CN102136140A (en) Rectangular pattern-based video image distance detecting method

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