CN107993249A - A kind of body gait data fusion method based on more Kinect - Google Patents
A kind of body gait data fusion method based on more Kinect Download PDFInfo
- Publication number
- CN107993249A CN107993249A CN201710727463.1A CN201710727463A CN107993249A CN 107993249 A CN107993249 A CN 107993249A CN 201710727463 A CN201710727463 A CN 201710727463A CN 107993249 A CN107993249 A CN 107993249A
- Authority
- CN
- China
- Prior art keywords
- kinect
- data
- skeleton
- method based
- data fusion
- 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
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
- G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/292—Multi-camera tracking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of body gait data fusion method based on more kinect, by building the data collecting system containing more kinect, it is synchronous to carry out gait data acquisition, and the initial data of collection is handled, including dry, back side knee joint data processing is gone in filtering;Then need to carry out the coordinate system of two Kinect unification, and then precision self compensation is carried out to the data of gained using bone precision curve, different convergence strategies is finally selected according to the difference of two Kinect placement positions, completes the gait data fusion of more Kinect.This method can solve the problems, such as how to extend collection apart from making up the shortcomings that Kinect detection ranges are limited, have successfully been obtained complete gait data.Meanwhile improve the data precision of collection;In addition, solve the problems, such as Kinect human body back data is gathered in knee joint data it is inaccurate, and employ opposition puts layout, that is, adds gait detecting distance, and the saving device space, improve practical application.
Description
One, technical fields
The present invention relates to the body gait Data Fusion method based on Kinect.
Two, background technologies
People investigated a large amount of interesting applications using Kinect in recent years, such as virtual fitting, video conference, face inspection
Survey, gesture identification, gait detection etc..The Kinect for windows SDK provided with reference to Microsoft, connect hardware device, just
The three-dimensional skeleton data of human body can be easily obtained by the interface that Microsoft provides.And Kinect can also directly acquire depth
Data, need troublesome configuration and complicated data processing, Kinect has more easily excellent compared to Conventional visual equipment
Gesture.
Because the limitation of hardware, a Kinect detection longest distance is 4.5m, and whole detecting distance is too short, it is impossible to complete
Human body walking data is presented, realizes data fusion currently with two Kinect, equipment can only as shown in Figure 1 put, easily make
Blocked into bone, influence left leg Data Detection, and expansion poor effect, can only achieve 6m or so.
The three, content of the invention
It is an object of the invention to provide a kind of three-dimensional data fusion method based on more Kinect, solve existing
Kinect distance and precision problem when extracting skeleton data, while solve skeleton data and saltus step occur or occur certainly
When blocking the problem of shortage of data.
In order to achieve the above object, the technical scheme is that at a kind of three-dimensional data fusion based on Kinect of design
Reason method, includes the following steps:
(1) horizontal opposition puts two Kinect, makes it towards shooting area, Kinect 1 and Kinect 2 respectively with electricity
Brain 4 and computer 5 connect, and LAN is established between computer 4 and computer 5, to carry out data transmission.
(2) C# programs are write under a windows environment, are obtained using bis- generations of vision collecting equipment Kinect under 10 human bodies
Limb skeletal joint point spatial data (X, Y, Z), corresponding joint are spinal roots, backbone middle part, left hip, left knee, a left side successively
Ankle, left foot, right hip, right knee, right ankle, right foot.
(3) median filter process is carried out to the initial data that step (2) obtains, filters out noise, the window length of medium filtering
Degree is definite according to Kinect message transmission rates (30 frame per second);
(4) when obtaining human body back side skeleton data, because having action of going down on one's knees in walking, depth number is obtained according to Kinect
According to principle understand, clothes has knee joint and blocks when going down on one's knees, and causes knee joint data to be not allowed, indirect using mathematics geometric method
Obtain accurate knee joint spatial data.
(5) left hip, left knee, two sets of public point coordinates of three artis of left ankle are taken, to the space coordinate of two Kinect
System carries out unification.
(6) hypotelorism due to human body to be measured and Kinect depth transducers, the error too far etc. caused by factor claim
For the geometric error of skeleton data.In order to eliminate this kind of error, it would be desirable to carry out precision self compensation to obtained skeleton data.
(7) data fusion is carried out for meeting the skeleton data of data fusion condition.
The advantages of the present invention are as follows:
(1) present invention proposes one kind and is based on Local Area Networt Communication Protocols, realizes that multiple stage computers connect Kinect respectively, together
Step carries out the new method of gait collection, coordinate unification, precision self compensation and data fusion, solves and how to extend collection distance
The problem of, the shortcomings that its detection range is limited is made up, have successfully been obtained complete gait data, gait cycle will be gathered from 3~4
It is a to increase to 8~9.
(2) present invention utilizes data fitting method, carries out precision self compensation to Kinect skeleton datas, improves collection
Precision.
(3) present invention solves the problems, such as that Kinect is inaccurate to knee joint data in the collection of human body back data, and uses
Opposition puts layout, that is, adds gait detecting distance, and saving device hardware space, improve practical application.
Four, are illustrated
Fig. 1 is that the more Kinect hardware of tradition put schematic diagram
Fig. 2 is somatic data fusion method flow chart
Fig. 3 is that somatic data fusion method hardware puts schematic diagram
Fig. 4 is Kinect space coordinate schematic diagrames
Fig. 5 is human body lower limbs joint index map
Fig. 6 is Strategy of data fusion classification chart
Five, embodiments (implementation of each several part)
A kind of three-dimensional data method for amalgamation processing based on more Kinect of the present invention, specifically implements, such as Fig. 2 according to the following steps
It is shown:
Step 1, data collecting system is built:As shown in figure 3, horizontal opposition puts two Kinect, make it towards shooting
Region 3, shooting area 3 determine by the camera watch region of two Kinect, Kinect 1 and Kinect 2 respectively with computer 4 and computer
5 connections, establish LAN, to carry out data transmission between computer 4 and computer 5.
Acquiescence computer 4 is primary processor, during system operation, personage's posture that computer 5 in real time can detect Kinect 2
Skeleton data is sent to computer 4 by LAN, while computer 4 can also obtain the posture skeleton data of Kinect 1 in real time.So
The position that computer 4 can be according to personage in area 3 afterwards, calls blending algorithm to carry out the data fusion of more Kinect.
Step 2, data acquisition:Kinect SDK are installed under windows platform, for vision collecting equipment Kinect and
Computer connects, and can smooth gathered data.C# scripts are write, obtain the sky of 10 human synovials of Kinect bone following functions
Between coordinate data, unit is rice, and coordinate origin is at depth camera, corresponding bone following function space coordinates such as Fig. 4 institutes
Show.What Kinect 1 was obtained is human body back side skeleton data, and what Kinect 2 was obtained is human body front skeleton data.
Wherein Fig. 5 be Kinect human body lower limbs joint index, be successively spinal roots, backbone middle part, left hip, left knee,
Left ankle, left foot, right hip, right knee, right ankle, right foot.Obtain in artis data procedures, it is desirable to which tester is immediately ahead of Kinect
In the range of 1.5-4.5m, it just can guarantee that the three dimensional space coordinate data in this 10 joints are effective.
Step 3, data filtering:According to gait initial data is obtained in step 2, noise is carried out using the method for medium filtering
Filter out.Medium filtering is suitable for the pulse interference signal accidentally occurred, is that a kind of theoretical based on sequencing statistical and effectively suppression is made an uproar
The nonlinear signal processing technology of sound.The basic principle of medium filtering is value present in digital signal sequences and its neighborhood
Each point value sequence, then goes the value of centre to replace the value of the point, so as to eliminate isolated noise spot.
, it is necessary to define one during medium filtering is carried out to 10 artis space coordinates (X, Y, Z) in step 2
The window of odd length 2n+1 (n is natural number).By taking hip joint X-coordinate value as an example, continuous 2n+1 X-coordinate value sequence is:X
(i-n) ..., X (i) ..., X (i+n), wherein X (i) obtain sampled value for knee joint X-coordinate value, which is pressed from small
To big sequence, result is after sequence:Sranged(1),Sranged(2),…,Sranged(2n+1), medium filtering output valve W are Sranged
(n+1), as shown in formula 1.
W=Sranged(n+1)=Med (Sranged(1),Sranged(2) ..., Sranged(2n+1)) (1)
Wherein, Med functions are the function for taking sequence median.
Step 4, back side knee joint data are handled:When obtaining human body back side skeleton data, because people walked normally
Cheng Zhong, has action of going down on one's knees, and trousers can be bent together with knee joint, according to Kinect obtains the principle of depth data,
What Kinect was detected is the depth data of trousers bent portion at this time, is not the position of human body knee joint articular cavity;That is trousers
Knee joint data are caused to be not allowed.We give up the inexact data measured, and knee joint space is obtained indirectly by mathematics geometric method
Coordinate.
Known hip joint space coordinate (XHip,YHip,ZHip), ankle-joint space coordinate (XAnkle,YAnkle,ZAnkle), thigh
Long LThight, leg length LCalf, knee joint X-axis coordinate XKnee, back side knee joint space coordinate can be solved by formula (2):
Step 5, two Kinect unified coordinate systems:Transformational relation between two coordinate systems is as follows,
Wherein, (X, Y, Z)TFor the coordinate value (X', Y', Z') measured under No. 1 KinectTFor the seat measured under No. 2 Kinect
Scale value, R are spin matrix (containing tri- parameters of α β γ) (Δ x, Δ y, Δ z)TRepresent translation parameters)
Common point (the X with two sets of joint coordinates provided hereini,Yi,Zi)、(Xi',Yi',Zi') i=1,2 ..., n n
≥3
For convenience of calculating, center of gravity processing is carried out to used coordinate, the coordinate of the common point of Two coordinate system is drawn
It is counted as the center of gravity coordinate using center of gravity as originThe barycentric coodinates of two other coordinate system are:(XR,
YR,ZR)、(X'R,YR',Z'R)
(4), which are substituted into (3), to be obtained:
Left hip, left knee, two sets of public point coordinates of three artis of left ankle are taken, spin matrix are obtained by (5) formula, then lead to
Cross (6) formula and obtain translation parameters.
Step 6, skeleton data precision self compensation:Due to the hypotelorism of human body to be measured and Kinect depth transducers, mistake
Error caused by the factor such as remote is known as the geometric error of skeleton data.In order to eliminate this kind of error, it would be desirable to what is obtained
Skeleton data carries out precision self compensation.
Kinect skeleton datas precision approximately linear relation in detection interval [1.5m-3m], at section [3m-4.5m]
On similar to polynomial of degree n functional relation, by fitting experimental data, obtain Kinect skeleton data trueness errors curve difference
For n(1,j)(l) and n(2,j)(l)。
Yj' (l)=Yj(l)-nj(l) j=1,2 ..., N
Wherein, Yj' (l) represent the data after the compensation of j-th joint, nj(l) true measurement Y is representedj(l) measurement on
Error.
Step 7, data fusion is carried out:Distance L is put to two Kinect to analyze, as shown in fig. 6, three kinds can be separated
Different convergence strategies, is L=6m, 6m respectively<L≤7.5m and 7.5m<L≤8.5m;Fusion under three kinds of convergence strategies is calculated
Method is classified according to skeleton data trueness error curve, can carry the obtained skeleton data precision in working region significantly
It is high.
1) for meeting the skeleton point of A={ only 1 gathered datas of Kinect }, Yj' (l)=Y('1,j)(l)
2) for meeting the skeleton point of B={ only 2 gathered datas of Kinect }, Yj' (l)=Y('2,j)(l)
3) for meeting the skeleton point of C={ Kinect 1 and Kinect 2 can gathered data }, Yj' (l)=(Y('1,j)
(l)+Y('2,j)(l))/2。
Wherein, Y('1,j)(l) data after j-th of the joint compensation of Kinect 1, Y are represented('2,j)(l) Kinect 2 is represented
Data after j-th of joint compensation.
Claims (5)
1. a kind of body gait data fusion method based on more Kinect, the main hardware system built including integration program,
The processing method of human body back skeleton data, the precision self compensation of Kinect skeleton datas and data fusion scheme.
2. a kind of body gait data fusion method based on more Kinect according to claim 1, it is characterised in that melt
The hardware of conjunction scheme includes two Kinect, two computers;The horizontal opposition of two Kinect is put, and distance is 6m≤L≤8.5m,
Two computers connect two Kinect respectively, and establish LAN each other, to carry out data transmission.
3. a kind of body gait data fusion method based on more Kinect according to claim 1, it is characterised in that
The hip joint space coordinate known, ankle-joint space coordinate, thigh length, leg length, knee joint X-axis coordinate, using mathematics it is several where
Method can solve back side knee joint space coordinate.
A kind of 4. body gait data fusion method based on more Kinect according to claim 1, it is characterised in that
Kinect skeleton datas precision approximately linear relation in detection interval [1.5m-3m], on section [3m-4.5m] similar to
Polynomial of degree n functional relation, by fitting experimental data, obtains Kinect skeleton data trueness error curves, bent using error
Line carries out precision self compensation to measurement data.
5. a kind of body gait data fusion method based on more Kinect according to claim 1, it is characterised in that 1)
Skeleton point for meeting A=only 1 gathered datas of Kinect, directly uses No. 1 Kinect skeleton data;2) for full
The skeleton point of sufficient B=only 2 gathered datas of Kinect, directly uses No. 2 Kinect skeleton datas;3) for meeting C=
The skeleton point of { Kinect 1 and Kinect 2 can gathered data }, using the average of two Kinect skeleton datas;Meanwhile
Skeleton point for meeting C, because self compensation curve and distance dependent, the difference of distance is put according to two Kinect, is also divided
The convergence strategy different into three kinds.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710727463.1A CN107993249A (en) | 2017-08-23 | 2017-08-23 | A kind of body gait data fusion method based on more Kinect |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710727463.1A CN107993249A (en) | 2017-08-23 | 2017-08-23 | A kind of body gait data fusion method based on more Kinect |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107993249A true CN107993249A (en) | 2018-05-04 |
Family
ID=62029805
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710727463.1A Pending CN107993249A (en) | 2017-08-23 | 2017-08-23 | A kind of body gait data fusion method based on more Kinect |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107993249A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108876881A (en) * | 2018-06-04 | 2018-11-23 | 浙江大学 | Figure self-adaptation three-dimensional virtual human model construction method and animation system based on Kinect |
CN109373993A (en) * | 2018-10-09 | 2019-02-22 | 深圳华侨城文化旅游科技股份有限公司 | A kind of positioning system and method based on more somatosensory devices |
CN111035393A (en) * | 2019-12-13 | 2020-04-21 | 中国科学院深圳先进技术研究院 | Three-dimensional gait data processing method, system, server and storage medium |
CN111461029A (en) * | 2020-04-03 | 2020-07-28 | 西安交通大学 | Human body joint point data optimization system and method based on multi-view Kinect |
CN111582081A (en) * | 2020-04-24 | 2020-08-25 | 西安交通大学 | Multi-Kinect serial gait data space-time combination method and measuring device |
CN111714129A (en) * | 2020-05-07 | 2020-09-29 | 广西科技大学 | Human gait information acquisition system |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1406215A1 (en) * | 2002-10-02 | 2004-04-07 | Université Libre de Bruxelles | Method for the modeling of skeletal kinematics |
CN104933734A (en) * | 2015-06-26 | 2015-09-23 | 西安理工大学 | Multi-Kinect-based human body gesture data fusion method |
-
2017
- 2017-08-23 CN CN201710727463.1A patent/CN107993249A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1406215A1 (en) * | 2002-10-02 | 2004-04-07 | Université Libre de Bruxelles | Method for the modeling of skeletal kinematics |
CN104933734A (en) * | 2015-06-26 | 2015-09-23 | 西安理工大学 | Multi-Kinect-based human body gesture data fusion method |
Non-Patent Citations (2)
Title |
---|
吴楠等: "基于多台Kinect传感器的步态采集***", 《生物医学工程研究》 * |
杜惠斌等: "基于集员滤波的双Kinect人体关节点数据融合", 《自动化学报》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108876881A (en) * | 2018-06-04 | 2018-11-23 | 浙江大学 | Figure self-adaptation three-dimensional virtual human model construction method and animation system based on Kinect |
CN109373993A (en) * | 2018-10-09 | 2019-02-22 | 深圳华侨城文化旅游科技股份有限公司 | A kind of positioning system and method based on more somatosensory devices |
CN111035393A (en) * | 2019-12-13 | 2020-04-21 | 中国科学院深圳先进技术研究院 | Three-dimensional gait data processing method, system, server and storage medium |
CN111035393B (en) * | 2019-12-13 | 2022-08-09 | 中国科学院深圳先进技术研究院 | Three-dimensional gait data processing method, system, server and storage medium |
CN111461029A (en) * | 2020-04-03 | 2020-07-28 | 西安交通大学 | Human body joint point data optimization system and method based on multi-view Kinect |
CN111461029B (en) * | 2020-04-03 | 2023-05-02 | 西安交通大学 | Human body joint point data optimization system and method based on multi-view Kinect |
CN111582081A (en) * | 2020-04-24 | 2020-08-25 | 西安交通大学 | Multi-Kinect serial gait data space-time combination method and measuring device |
CN111714129A (en) * | 2020-05-07 | 2020-09-29 | 广西科技大学 | Human gait information acquisition system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107993249A (en) | A kind of body gait data fusion method based on more Kinect | |
Zago et al. | 3D tracking of human motion using visual skeletonization and stereoscopic vision | |
CN110599540B (en) | Real-time three-dimensional human body shape and posture reconstruction method and device under multi-viewpoint camera | |
Schreven et al. | Optimising filtering parameters for a 3D motion analysis system | |
CN111931733B (en) | Human body posture detection method based on depth camera | |
CN107578019B (en) | Gait recognition system and method based on visual sense and tactile sense fusion | |
CN110801226A (en) | Human knee joint moment testing system method based on surface electromyographic signals and application | |
CN108022248A (en) | A kind of lower limb gait rehabilitation assessment system of view-based access control model collecting device | |
CN104408718B (en) | A kind of gait data processing method based on Binocular vision photogrammetry | |
CN103955316B (en) | A kind of finger tip touching detecting system and method | |
CN111383221B (en) | Scoliosis detection model generation method and computer equipment | |
CN104035557B (en) | Kinect action identification method based on joint activeness | |
CN110176062A (en) | A kind of control method of the virtual human body based on movement capturing data | |
WO2023273093A1 (en) | Human body three-dimensional model acquisition method and apparatus, intelligent terminal, and storage medium | |
CN104167016A (en) | Three-dimensional motion reconstruction method based on RGB color and depth image | |
CN107049270A (en) | System and method for measuring pulse wave translation time | |
CN113856186B (en) | Pull-up action judging and counting method, system and device | |
CN113268141B (en) | Motion capture method and device based on inertial sensor and fabric electronics | |
CN102932638A (en) | 3D video monitoring method based on computer modeling | |
WO2020147791A1 (en) | Image processing method and device, image apparatus, and storage medium | |
CN102217952B (en) | Vector loop diagram generation method and device based on myocardium movement locus | |
CN107336233B (en) | Inertial-kinetic-capture-based human-robot virtual-real interaction control system | |
CN109191559A (en) | Manikin method for building up based on depth transducer | |
CN205850175U (en) | Motion correcting instrument | |
CN111312363A (en) | Double-hand coordination enhancing system based on virtual reality |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180504 |
|
WD01 | Invention patent application deemed withdrawn after publication |