CN107481286A - Dynamic 3 D schematic capture algorithm based on passive infrared reflection - Google Patents
Dynamic 3 D schematic capture algorithm based on passive infrared reflection Download PDFInfo
- Publication number
- CN107481286A CN107481286A CN201710562154.3A CN201710562154A CN107481286A CN 107481286 A CN107481286 A CN 107481286A CN 201710562154 A CN201710562154 A CN 201710562154A CN 107481286 A CN107481286 A CN 107481286A
- Authority
- CN
- China
- Prior art keywords
- point
- human body
- apar
- mark
- bands
- 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
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/254—Analysis of motion involving subtraction of images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Image Analysis (AREA)
Abstract
The present invention proposes the dynamic 3 D schematic capture algorithm based on passive infrared reflection, is related to a kind of seizure algorithm, comprises the following steps that:1st, manikin is established;2nd, human body sketch figure picture is obtained by background subtraction;3rd, the characteristic point on human body contour outline is extracted;4th, apar bands are detected and identify each mark point;5th, Kalman filter, epipolar-line constraint and human body are constrained to track identified mark point;6th, three-dimensional reconstruction is carried out so as to complete human body motion capture.The present invention is caught using passive infrared reflection technology to human motion, human body sketch figure picture is extracted with background subtraction, human region is detected with apar bands, and combine the position of mark point and number on human body and each mark point is identified, then the constraint of the mark point Kalman filter in the image of each frame, epipolar-line constraint and human body is tracked, finally the three-dimensional space position of mark point is carried out with the method for binocular stereo vision to rebuild the purpose for reaching human body motion capture.
Description
Technical field
The present invention relates to one kind to catch algorithm, and especially a kind of dynamic 3 D schematic capture based on passive infrared reflection is calculated
Method.
Background technology
Human body motion capture is the computer technology to receive much concern in recent years, and it is in intelligent monitoring, man-machine interaction, motion point
The fields such as analysis have broad application prospects.Human body motion capture refers to recovers human motion from one or more image sequences
Process, human motion here is the motion of finger, four limbs, trunk, not including the small yardstick action such as expression and gesture.
Conventional capturing technology can be divided into mechanical, acoustics formula, electromagnetic type and optical profile type from the principle.And independent of
Sensor special, the movement capturing technology of Direct Recognition characteristics of human body will also move towards practical quickly.The equipment of different principle respectively has
Its advantage and disadvantage, it can typically be evaluated from the following aspects:Positioning precision, real-time, degree easy to use, motion can be caught
Range size, cost, anti-interference and multiple target capturing ability.For the angle of technology, the essence of seizure seek to measurement,
The movement locus of tracking and record object in three dimensions.Typical motion capture device is typically by following components group
Into:Sensor, signal capture equipment, data transmission set and data processing equipment.
It is how clear using the sensing element of high sensitivity, intake for being applied to human body seizure system in practice
Clear image, to complete human body dynamic capture system by simple information processing system be that those skilled in the art need to solve
Problem.
The content of the invention
The present invention provides a kind of dynamic 3 D schematic capture algorithm based on passive infrared reflection, accurately catches three-dimensional structure
Action.
The present invention specifically adopts the following technical scheme that realization:
A kind of dynamic 3 D schematic capture algorithm based on passive infrared reflection, is comprised the following steps that:
Step 1, establish manikin;
Step 2, pass through background subtraction acquisition human body sketch figure picture;
Characteristic point in step 3, extraction human body contour outline;
Step 4, detection apar bands simultaneously identify each mark point;
Step 5, Kalman filter, epipolar-line constraint and human body are constrained to track identified mark point;
Step 6, three-dimensional reconstruction is carried out so as to complete human body motion capture.
Preferably, the head of the manikin, trunk and four limbs are made up of six apar bands, and mark point is distinguished
It is attached to forehead, wrist, ancon, shoulder, hip, knee, ankle.
Preferably, the mark point uses infrared reflecting film as mark point.
The onestep extraction preferably, flex point that the step 3 is detected on human body contour outline according to the change in chain code direction is gone forward side by side
Go out characteristic point, comprise the following steps that:
Step 31, the starting point of note chain code are O, and travel through each point on profile.If the direction of some point uplink code
Change, then it is checked to starting point O distance (distance refer to put between the point and a upper flex point on profile number),
If distance is less than threshold value Tl, continue to travel through the point of next chain code direction change until some point to starting point O distance are big
In threshold value Tl, this, which is put, is referred to as flex point, repeats above-mentioned sampling process until having traveled through the institute on human body contour outline a little;
Step 32, select characteristic point followed by the traversal obtained flex point of sampling:If current flex point is P1, it is previous
Flex point is P0, the latter flex point is P2, calculate vector (P0, P1) and vector (P1, P2) between angle theta, if θ is more than threshold value
Tθ, then flex point P1As characteristic point.Further, since human body stand when, trunk profile may with leg lateral profile almost
On a line, on such profile from oxter to ankle in any point will not be detected as characteristic point, this can cause nothing
Method accomplishes the identification of trunk and leg, therefore, during actually detected, flex point is sampled when one from previous characteristic point
Distance is more than TmWhen, it is also considered as characteristic point.
Preferably, the step 4 connects the characteristic point of extraction using real segment to be fitted human body wheel counterclockwise
Exterior feature, comprise the following steps that:
If step 41, the length-width ratio of apar bands are close to 1.0 and only one of which mark point, then the apar bands
It is exactly the mark point on head, mark point i.e. crown position;
If step 42, the length-width ratio of apar bands close to 1.0 and without mark point or have two mark points, then
The apar bands be exactly the trunk of human body and the two marks should be hip mark, these mark points are judging leg
It can also be identified during apar bands;
Two apar bands that if there are 3 mark points step 43, some apar band the inside and it is neighbouring are head respectively
Portion and trunk, then the apar bands are considered as arm, and left arm or right arm are determined relative to the position on head according to it,
Mark point close to apar band blind ends is wrist, and the other end is shoulder, and what it is in centre is then ancon;
Step 44, remaining two apar bands are then legs, and between which also has 3 mark points.According to it relative to head
The position in portion determines left leg or right leg, and the mark point close to apar band blind ends is ankle, the other end is hip,
What it is in centre is then knee.
Dynamic 3 D schematic capture algorithm provided by the invention based on passive infrared reflection, its advantage are:Profit
Human motion is caught with passive infrared reflection technology, extracts human body sketch figure picture, Ran Houyong with background subtraction first
Apar bands detect human region, including head, trunk and four limbs, and combine on human body the position of mark point and number to each mark
Note point identified, then to the mark point in the image of each frame with Kalman filter, epipolar-line constraint and human body constrain into
Line trace, finally the three-dimensional space position of mark point is rebuild so as to reach human motion with the method for binocular stereo vision
The purpose of seizure.
Brief description of the drawings
Fig. 1 is the flow chart of dynamic 3 D schematic capture algorithm of the present invention;
Fig. 2 is the schematic diagram of manikin;
Fig. 3 is background image schematic diagram;
Fig. 4 is input picture schematic diagram;
Fig. 5 is human body sketch figure picture schematic diagram;
Fig. 6 is human body contour outline image schematic diagram;
Fig. 7 is to extract the characteristic point schematic diagram on arm;
Fig. 8 is the characteristic point schematic diagram on profile;
Fig. 9 is dual camera observation space point P schematic diagrames.
Embodiment
To further illustrate each embodiment, the present invention is provided with accompanying drawing.These accompanying drawings are the invention discloses the one of content
Point, it can coordinate the associated description of specification to explain the operation principles of embodiment mainly to illustrate embodiment.Coordinate ginseng
These contents are examined, those of ordinary skill in the art will be understood that other possible embodiments and advantages of the present invention.In figure
Component be not necessarily to scale, and similar element numbers are conventionally used to indicate similar component.
In conjunction with the drawings and specific embodiments, the present invention is further described.
As shown in figure 1, the dynamic 3 D schematic capture algorithm based on passive infrared reflection that the present embodiment provides, is divided into two
Individual part:Initialization and tracking.Initialization section uses the position for identifying each mark point on body.Initially set up human body
Model, human body sketch figure picture is obtained by background subtraction, then extracts the characteristic point on human body contour outline, and combine human region
Feature recognition mark point.Tracking section is constrained with Kalman filter, epipolar-line constraint and human body to track identified mark point
And three-dimensional reconstruction is carried out so as to complete human body motion capture.Algorithm detailed step is as follows:
Step 1, as shown in Fig. 2 the manikin of dynamic 3 D schematic capture, head, trunk and the four limbs in figure are by six
Individual apar bands composition, and stain represents mark point in figure, be attached to respectively forehead, wrist, ancon, shoulder, hip, knee,
These feature locations of ankle, for the accurate position for calculating human synovial, the mark point of the present embodiment uses infrared reflecting film
As mark point, infrared reflecting film is a kind of material to reflection infrared light at infrared light supply, uses the good of this reflective membrane
As long as place is that camera and infrared transmitting tube are put together, camera is just bound to capture infrared light, and so doing will not be to
User brings any extra bear a heavy burden and simple and convenient.Solid line represents the length of human limb in figure, and they are in theory
Constant, this can be used as the tracking that human body constraint carrys out aid mark point.
Step 2, background subtraction is subtracted with present image with reference to background model to realize extraction foreground object, the party
Method is generally possible to provide very safe characteristic, the foreground object in can more fully extracting in image, position it is accurate and
Speed is fast.Fig. 3 is the background image that camera photographs, and brightness is very low, it can be seen that ground and object above are because catching
All bound infrared auxiliary lamp on two cameras of human motion, thus there is a small amount of infrared light to be diffusely reflected and by camera
Photographed.When user enters in the range of shooting, the image of shooting is as shown in figure 4, due to nearer from camera, with user
Diffusing reflection effect is relatively good, it is possible to human body is seen from figure, although also there is shade on ground, due to the brightness on ground
It is inherently very low, cover top shadow after brightness vary less, by selecting appropriate threshold value to eliminate the interference of shade.If clap
The background image taken the photograph is B, and first image shot during user's formal training is f, then
T is threshold value, and it is the bianry image after background subtraction to be taken as 5, D.Due to the interference of noise, aperture occurs in image D
Hole and burr, it can fill these cavities using image in morphological image expansion and etching operation and remove burr,
The largest connected region finally obtained is exactly the sketch figure picture of human body, as shown in figure 5, the hair of people also brightness and background phase in figure
Poor very little and be removed, but whole contouring head can be replaced with face.
Step 3, after sketch figure picture is obtained, detect the profile of human body and represented with the form of Freeman chain codes, human body wheel
Wide image is as shown in Figure 6.Characteristic point is gone out according to the flex point onestep extraction of going forward side by side on the change detection human body contour outline in chain code direction, walked
It is rapid as follows:
Step 31, the starting point of note chain code are O, and travel through each point on profile.If the direction of some point uplink code
Change, then it is checked to starting point O distance (distance refer to put between the point and a upper flex point on profile number),
If distance is less than threshold value Tl, continue to travel through the point of next chain code direction change until some point to starting point O distance are big
In threshold value Tl, this, which is put, is referred to as flex point.Above-mentioned sampling process is repeated until having traveled through the institute on human body contour outline a little.
Step 32, select characteristic point followed by the traversal obtained flex point of sampling:If current flex point is P1, it is previous
Flex point is P0, the latter flex point is P2, calculate vector (P0, P1) and vector (P1, P2) between angle theta, as shown in Figure 7.If θ
More than threshold value Tθ, then flex point P1As characteristic point.Further, since when human body is stood, trunk profile may be with the outside of leg
Profile almost on a line, on such profile from oxter to ankle in any point will not be detected as characteristic point, this
The identification for accomplishing trunk and leg can be led to not.Therefore, during actually detected, flex point is sampled when one from previous
The distance of characteristic point is more than TmWhen, it is also considered as characteristic point.
Threshold value Tl and T θ are used for the density of controlling feature point, if they are too small, will obtain the characteristic point of too many redundancy;
On the other hand, if they are too big, degree of fitting step-down of these characteristic points to human body contour outline can be caused again.In the realization of the system
In, T is setlFor 30, TθFor 20 °, so that the number of characteristic point is controlled between 20 to 30, and TmSetting with people
Height on image is relevant, and 1/4, the Fig. 8 for being typically set to pixel distance from top to bottom shows the extraction result of characteristic point.
Step 4, apar bands are detected and identify each mark point, connect the feature of extraction counterclockwise using real segment
Put to be fitted human body contour outline, each line segment has direction, i.e., points to later feature point from previous characteristic point.These have direction
Line segment be used to detect apar bands, comprise the following steps that:
If step 41, the length-width ratio of apar bands are close to 1.0 and only one of which mark point, then the apar bands
It is exactly the mark point on head, mark point i.e. crown position;
If step 42, the length-width ratio of apar bands close to 1.0 and without mark point or have two mark points, then
The apar bands be exactly the trunk of human body and the two marks should be hip mark, these mark points are judging leg
It can also be identified during apar bands;
Two apar bands that if there are 3 mark points step 43, some apar band the inside and it is neighbouring are head respectively
Portion and trunk, then the apar bands are considered as arm, and left arm or right arm are determined relative to the position on head according to it,
Mark point close to apar band blind ends is wrist, and the other end is shoulder, and what it is in centre is then ancon;
Step 44, remaining two apar bands are then legs, and between which also has 3 mark points.According to it relative to head
The position in portion determines left leg or right leg, and the mark point close to apar band blind ends is ankle, the other end is hip,
What it is in centre is then knee.
Step 5, Kalman filter, after the predicted position of mark point in the current frame is obtained, using the position of prediction in
The heart search length of side is the square-shaped frame of 15 pixels to detect in rectangle whether have mark point.Marked when there was only one in prediction block
During point, in this case, each variable of Kalman filter is directly updated according to the center of mark point in prediction block
Value.When there is no mark point in the frame of prediction, if all mark points detected have all matched with other joint, then just
The position of prediction is updated into each variate-value of Kalman filter as the position of detection;If also mark point does not match,
Then these mark points are matched with the joint not matched according to epipolar-line constraint and human body constraint.Have inside prediction block multiple
During mark point, also to differentiate each mark point should match with which joint according to epipolar-line constraint and human body constraint, obtain just
After true result, then update each variate-value of Kalman filter.
Step 6, the three-dimensional space position for calculating mark point, such as Fig. 9, for any point P in three dimensions, if only used
One camera C1To observe it, then it is seen that point P is in camera C1Image point p on the image of shooting1, pass through the image point
The positions of point P in three dimensions can not be obtained.In fact, ray O1P(O1For camera C1Photocentre) on any point P
' in camera C1Image point on the image of shooting is all p1, so by image point p1Position point position in space can only be determined in ray
O1p1On, that is, the depth of spatial point can not be determined.If with two camera C1And C2Carry out point of observation P, it is in camera C1
Image point on the image of shooting is p1, in camera C2Image point on the image of shooting is p2, then spatial point P is both located at ray
O1p1On, it is located at ray O again2p2On, therefore point P1Exactly ray O1p1And O2p2Intersection point, this sampling point P spatial three-dimensional position is just
Uniquely determine.
The application is caught using passive infrared reflection technology to human motion, extracts human body with background subtraction first
Sketch figure picture, human region, including head, trunk and four limbs then are detected with apar bands, and combine the position of mark point on human body
Put and each mark point is identified with number, then to the mark point Kalman filter in the image of each frame, polar curve
Constraint and human body constraint are tracked, and finally the three-dimensional space position of mark point is rebuild with the method for binocular stereo vision
So as to reach the purpose of human body motion capture.
Although specifically showing and describing the present invention with reference to preferred embodiment, those skilled in the art should be bright
In vain, do not departing from the spirit and scope of the present invention that appended claims are limited, in the form and details can be right
The present invention makes a variety of changes, and is protection scope of the present invention.
Claims (5)
1. a kind of dynamic 3 D schematic capture algorithm based on passive infrared reflection, it is characterised in that comprise the following steps that:
Step 1, establish manikin;
Step 2, pass through background subtraction acquisition human body sketch figure picture;
Characteristic point in step 3, extraction human body contour outline;
Step 4, detection apar bands simultaneously identify each mark point;
Step 5, Kalman filter, epipolar-line constraint and human body are constrained to track identified mark point;
Step 6, three-dimensional reconstruction is carried out so as to complete human body motion capture.
2. the dynamic 3 D schematic capture algorithm according to claim 1 based on passive infrared reflection, it is characterised in that:Institute
Head, trunk and the four limbs for stating manikin are made up of six apar bands, and mark point is attached to forehead, wrist, elbow respectively
Portion, shoulder, hip, knee, ankle.
3. the dynamic 3 D schematic capture algorithm according to claim 2 based on passive infrared reflection, it is characterised in that:Institute
Mark point is stated using infrared reflecting film as mark point.
4. the dynamic 3 D schematic capture algorithm according to claim 1 based on passive infrared reflection, it is characterised in that:Institute
State step 3 and characteristic point is gone out according to the flex point onestep extraction of going forward side by side on the change detection human body contour outline in chain code direction, specific steps are such as
Under:
Step 31, the starting point of note chain code are O, and travel through each point on profile.If the direction of some point uplink code occurs
Change, then it is checked to starting point O distance (distance refer to put between the point and a upper flex point on profile number), if
Distance is less than threshold value Tl, continue to travel through the point of next chain code direction change until some point to starting point O distance are more than threshold
Value Tl, this, which is put, is referred to as flex point, repeats above-mentioned sampling process until having traveled through the institute on human body contour outline a little;
Step 32, select characteristic point followed by the traversal obtained flex point of sampling:If current flex point is P1, previous flex point
For P0, the latter flex point is P2, calculate vector (P0, P1) and vector (P1, P2) between angle theta, if θ is more than threshold value Tθ, then
Flex point P1As characteristic point.Further, since human body stand when, trunk profile may with leg lateral profile almost one
On the line of bar one, on such profile from oxter to ankle in any point will not be detected as characteristic point, this can lead to not do
To the identification of trunk and leg, therefore, during actually detected, with a distance from a sampling flex point is from previous characteristic point
More than TmWhen, it is also considered as characteristic point.
5. the dynamic 3 D schematic capture algorithm according to claim 1 based on passive infrared reflection, it is characterised in that:Institute
State step 4 and connect the characteristic point of extraction counterclockwise using real segment to be fitted human body contour outline, comprise the following steps that:
If step 41, the length-width ratio of apar bands are close to 1.0 and only one of which mark point, then the apar bands are exactly
The mark point at head, mark point i.e. crown position;
If step 42, the length-width ratio of apar bands close to 1.0 and without mark point or have two mark points, then should
Apar bands be exactly the trunk of human body and the two marks should be hip mark, these mark points are judging the apar of leg
It can also be identified during band;
If have 3 mark points inside step 43, some apar band and two apar bands that it is neighbouring be respectively head and
Trunk, then the apar bands are considered as arm, and left arm or right arm are determined relative to the position on head according to it, are approached
The mark point of apar band blind ends is wrist, and the other end is shoulder, and what it is in centre is then ancon;
Step 44, remaining two apar bands are then legs, and between which also has 3 mark points.According to it relative to head
Position determines left leg or right leg, and the mark point close to apar band blind ends is ankle, the other end is hip, in
Between be then knee.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710562154.3A CN107481286A (en) | 2017-07-11 | 2017-07-11 | Dynamic 3 D schematic capture algorithm based on passive infrared reflection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710562154.3A CN107481286A (en) | 2017-07-11 | 2017-07-11 | Dynamic 3 D schematic capture algorithm based on passive infrared reflection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107481286A true CN107481286A (en) | 2017-12-15 |
Family
ID=60595071
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710562154.3A Pending CN107481286A (en) | 2017-07-11 | 2017-07-11 | Dynamic 3 D schematic capture algorithm based on passive infrared reflection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107481286A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108344738A (en) * | 2018-01-22 | 2018-07-31 | 翰飞骏德(北京)医疗科技有限公司 | Imaging method and its device for hydroxyapatite |
CN111562843A (en) * | 2020-04-29 | 2020-08-21 | 广州美术学院 | Positioning method, device, equipment and storage medium for gesture capture |
CN113688683A (en) * | 2021-07-23 | 2021-11-23 | 网易(杭州)网络有限公司 | Optical motion capture data processing method, model training method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800126A (en) * | 2012-07-04 | 2012-11-28 | 浙江大学 | Method for recovering real-time three-dimensional body posture based on multimodal fusion |
CN103198492A (en) * | 2013-03-28 | 2013-07-10 | 沈阳航空航天大学 | Human motion capture method |
CN106250867A (en) * | 2016-08-12 | 2016-12-21 | 南京华捷艾米软件科技有限公司 | A kind of skeleton based on depth data follows the tracks of the implementation method of system |
-
2017
- 2017-07-11 CN CN201710562154.3A patent/CN107481286A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800126A (en) * | 2012-07-04 | 2012-11-28 | 浙江大学 | Method for recovering real-time three-dimensional body posture based on multimodal fusion |
CN103198492A (en) * | 2013-03-28 | 2013-07-10 | 沈阳航空航天大学 | Human motion capture method |
CN106250867A (en) * | 2016-08-12 | 2016-12-21 | 南京华捷艾米软件科技有限公司 | A kind of skeleton based on depth data follows the tracks of the implementation method of system |
Non-Patent Citations (1)
Title |
---|
朱健: "沉浸式虚拟射击训练***中人机交互的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108344738A (en) * | 2018-01-22 | 2018-07-31 | 翰飞骏德(北京)医疗科技有限公司 | Imaging method and its device for hydroxyapatite |
CN111562843A (en) * | 2020-04-29 | 2020-08-21 | 广州美术学院 | Positioning method, device, equipment and storage medium for gesture capture |
CN113688683A (en) * | 2021-07-23 | 2021-11-23 | 网易(杭州)网络有限公司 | Optical motion capture data processing method, model training method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109059895B (en) | Multi-mode indoor distance measurement and positioning method based on mobile phone camera and sensor | |
CN102982557B (en) | Method for processing space hand signal gesture command based on depth camera | |
US9424679B2 (en) | Three-dimensional motion capture | |
US7239718B2 (en) | Apparatus and method for high-speed marker-free motion capture | |
US9330307B2 (en) | Learning based estimation of hand and finger pose | |
CN105956586B (en) | A kind of intelligent tracking system based on TOF 3D video camera | |
CN103279186B (en) | Merge the multiple goal motion capture system of optical alignment and inertia sensing | |
CN109299659A (en) | A kind of human posture recognition method and system based on RGB camera and deep learning | |
CN111353355B (en) | Motion tracking system and method | |
CN104408718B (en) | A kind of gait data processing method based on Binocular vision photogrammetry | |
CN107481286A (en) | Dynamic 3 D schematic capture algorithm based on passive infrared reflection | |
CN101930628A (en) | Monocular-camera and multiplane mirror catadioptric device-based motion capturing method | |
D'Eusanio et al. | Refinet: 3d human pose refinement with depth maps | |
CN115909396A (en) | Dynamic target tracking method for foot type robot | |
CN117824624A (en) | Indoor tracking and positioning method, system and storage medium based on face recognition | |
Zhou et al. | Information-efficient 3-D visual SLAM for unstructured domains | |
CN113487674A (en) | Human body pose estimation system and method | |
Yang et al. | Dance Posture Analysis Based on Virtual Reality Technology and Its Application in Dance Teac. | |
Shere et al. | Temporally consistent 3D human pose estimation using dual 360deg cameras | |
CN113421286B (en) | Motion capturing system and method | |
Zhuang et al. | Video motion capture using feature tracking and skeleton reconstruction | |
Abraham et al. | 3D ranging and tracking using lensless smart sensors | |
Ababsa | Advanced 3D localization by fusing measurements from GPS, inertial and vision sensors | |
JP6944144B2 (en) | Swing analyzer, method and program | |
Janardhanan et al. | A comprehensive study on human pose estimation |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171215 |