CN106096518A - Quick dynamic human body action extraction based on degree of depth study, recognition methods - Google Patents
Quick dynamic human body action extraction based on degree of depth study, recognition methods Download PDFInfo
- Publication number
- CN106096518A CN106096518A CN201610387248.7A CN201610387248A CN106096518A CN 106096518 A CN106096518 A CN 106096518A CN 201610387248 A CN201610387248 A CN 201610387248A CN 106096518 A CN106096518 A CN 106096518A
- Authority
- CN
- China
- Prior art keywords
- human body
- degree
- action
- recognition methods
- target
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of quick dynamic human body action extraction based on degree of depth study, recognition methods.Current existing human body recognition technology and application have the deficiency of the following aspects, a human skeleton inherently complicated structure, the motor habit of different people, and its manner of execution is the most different, and this has resulted in the difficulty identifying human body universality.The present invention comprises the steps: it is first to describe the size of human body target, color, edge, profile, shape and the Global Information of the degree of depth, it is provided with by clue for action recognition, effective motion feature is extracted from video sequence, in the case of distant view, the movement locus of target is utilized to carry out trajectory analysis；In the case of close shot, then need to utilize the information extracted from image sequence that the extremity of target and trunk are carried out the modeling of two dimension or three-dimensional.The present invention is for quick dynamic human body action extraction based on degree of depth study, recognition methods.
Description
Technical field:
The present invention relates to action recognition field, be specifically related to a kind of quick dynamic human body action based on degree of depth study and extract, know
Other method.
Background technology:
Current existing human body recognition technology and application have the deficiency of the following aspects, a human skeleton inherently complexity
Structure, the motor habit of different people, its manner of execution is the most different, this resulted in identify human body universality difficulty.Its
Secondary identified target location limitation is with mandatory, and identified target must adjust the position of oneself so that it is front alignment photographic head,
Side all may None-identified, furthermore the speed of being in response to and efficiency, for continuous print human action, the data before frame and frame are deposited
In redundancy, the memory space not only accounted for is big, also add amount of calculation.
Action recognition technology is motion characteristic based on people, the human body video flowing to input, first determines whether whether it exists
The mankind, if there is the mankind, the most further provide the position letter of everyone position, size and each major joint index point
Breath, and according to these information, extract the motion characteristic contained in everyone action further, and it is special with known action
Levying and contrast, thus identify everyone ongoing action, action recognition technology has a wide range of applications, such as dangerous play
Identify warning, human-computer interaction, medical assistance behavior modification, production of film and TV etc..
In order to quickly obtain and identify human body behavior, overcome above-mentioned the deficiencies in the prior art, the invention provides one
The recognition methods learnt based on the computer degree of depth, that the method can be real-time, dynamic, quick, extraction and knowledge on a large scale
Others' body action, can preferably be applied to the systems such as enterprise, school, the dangerous play alarm of government bodies, production of film and TV, fall
Low hsrdware requirements, traditional method is quite strict, in the identification to ten thousand people's ranks, commonly for hsrdware requirements in action recognition
Computer cannot meet computing demand at all, and this method passes through single knuckle, and double floating-point operations effectively solve hardware and took
Big problem, randomly draws in filler test 100,000 people, only needs the identification of people extremely short to accomplish Real time identification.
Summary of the invention:
It is an object of the invention to provide a kind of quick dynamic human body action extraction based on degree of depth study, recognition methods.
Above-mentioned purpose is realized by following technical scheme:
A kind of based on the degree of depth study quick dynamic human body action extraction, recognition methods, first be describe human body target size,
The Global Information of color, edge, profile, shape and the degree of depth, is provided with by clue for action recognition, extracts from video sequence
Effective motion feature, in the case of distant view, utilizes the movement locus of target to carry out trajectory analysis；In the case of close shot, then need profit
By the information extracted from image sequence, the extremity of target and trunk are carried out the modeling of two dimension or three-dimensional.
Described quick dynamic human body action extraction based on degree of depth study, recognition methods, by search body dimension, face
The features such as color, edge, profile, shape, determine that mobile object is the mankind, intercept human body image by screening afterwards, then people
With it on major joint position or more position, the index point that can be identified and follow the trail of is set, is shot by video camera
The action of same human body, then according to Space geometric parameter, in conjunction with the mathematical model of some human motions, can extrapolate people
The most each index point is in the position in each moment, and the combination of multiple index point positions just constitutes the integral position of human body, enters
Row continuous print location recognition thus identify human action.
Described quick dynamic human body action extraction based on degree of depth study, recognition methods, to taking the photograph of described video camera
As the image of head collection processes in real time, first gather and tell the people in image, pedestrian region frame is elected, so
After each frame in this region is contrasted with its former frame and a later frame, calculate the mobile change of pixel in three frames, logical
Cross and calculate the OpticalFlow that moves of pixel, there is shown the motion vector (Fx, Fy) that pixel moves, it follows that by this to
Amount is decomposed so that:, after Gaussian filter filters, just obtain the character representation of paid close attention to pedestrian's action.
Beneficial effect:
1. the present invention is a kind of quick dynamic human body action extraction based on degree of depth study, recognition methods, mainly provides one
Kind of recognition methods based on the study of the computer degree of depth, that the method can be real-time, dynamic, quickly, on a large scale extract and
Identify human action, can preferably be applied to the systems such as enterprise, school, the dangerous play alarm of government bodies, production of film and TV,
Reducing hsrdware requirements, traditional method is quite strict for hsrdware requirements in action recognition, in the identification to ten thousand people's ranks, general
Logical computer cannot meet computing demand at all.
The present invention pass through single knuckle, double floating-point operations effectively solve hardware and take excessive problem, 100,000 people with
In machine extraction filler test, only need the identification of people extremely short to accomplish Real time identification.
Accompanying drawing illustrates:
Accompanying drawing 1 is the action recognition technology schematic diagram of the neutral net of the present invention.
Accompanying drawing 2 is the motion characteristic figure of the present invention.
Detailed description of the invention:
Embodiment 1:
A kind of based on the degree of depth study quick dynamic human body action extraction, recognition methods, first be describe human body target size,
The Global Information of color, edge, profile, shape and the degree of depth, is provided with by clue for action recognition, extracts from video sequence
Effective motion feature, in the case of distant view, utilizes the movement locus of target to carry out trajectory analysis；In the case of close shot, then need profit
By the information extracted from image sequence, the extremity of target and trunk are carried out the modeling of two dimension or three-dimensional.
Embodiment 2:
According to the quick dynamic human body action extraction based on degree of depth study described in embodiment 1, recognition methods, by search human body
The features such as size, color, edge, profile, shape, determine that mobile object is the mankind, intercept human body image by screening afterwards, so
After on the person, on major joint position or more position, the index point that can be identified and follow the trail of is set, by shooting
Machine shoots the action of same human body, then according to Space geometric parameter, in conjunction with the mathematical model of some human motions, can push away
Calculating on the person that each index point is in the position in each moment, the combination of multiple index point positions just constitutes the overall position of human body
Put, carry out continuous print location recognition thus identify human action.
Embodiment 3:
According to the quick dynamic human body action extraction based on degree of depth study described in embodiment 1 or 2, recognition methods, to described
The image of the camera collection of video camera processes in real time, first gathers and tells the people in image, by pedestrian region
Frame is elected, and is then contrasted with its former frame and a later frame by each frame in this region, calculates pixel in three frames
Mobile change, by calculating the OpticalFlow that pixel moves, there is shown the motion vector (Fx, Fy) that pixel moves, connects
Get off, this vector decomposed so that:, after Gaussian filter filters, just obtain the mark sheet of paid close attention to pedestrian's action
Show；
Motion characteristic image sees accompanying drawing 2
Feature calculation formula:
。
Claims (3)
1. quick dynamic human body action extraction based on degree of depth study, a recognition methods, is characterized in that: be first to describe human body
The Global Information of the size of target, color, edge, profile, shape and the degree of depth, is provided with by clue for action recognition, from video
Sequence extracts effective motion feature, in the case of distant view, utilizes the movement locus of target to carry out trajectory analysis；Close shot feelings
Under condition, then need to utilize the information extracted from image sequence that the extremity of target and trunk are carried out the modeling of two dimension or three-dimensional.
Quick dynamic human body action extraction based on degree of depth study the most according to claim 1, recognition methods, its feature
It is: by features such as search body dimension, color, edge, profile, shapes, determine that mobile object is the mankind, afterwards by screening
Intercept human body image, then arrange on major joint position or more position on the person and can be identified and follow the trail of
Index point, shoots the action of same human body, then according to Space geometric parameter, in conjunction with some human motions by video camera
Mathematical model, can extrapolate on the person that each index point is in the position in each moment, and the combination of multiple index point positions is with regard to structure
Become the integral position of human body, carried out continuous print location recognition thus identify human action.
Quick dynamic human body action extraction based on degree of depth study the most according to claim 1 and 2, recognition methods, it is special
Levy and be: the image of the camera collection of described video camera is processed in real time, first gather and tell the people in image, will
Pedestrian region frame is elected, and is then contrasted with its former frame and a later frame by each frame in this region, calculates three
The mobile change of pixel in frame, by calculating the OpticalFlow that pixel moves, there is shown the motion vector that pixel moves
(Fx, Fy), it follows that decompose this vector so that:, through gaussian filtering
After device filtering, just obtain the character representation of paid close attention to pedestrian's action.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610387248.7A CN106096518A (en) | 2016-06-02 | 2016-06-02 | Quick dynamic human body action extraction based on degree of depth study, recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610387248.7A CN106096518A (en) | 2016-06-02 | 2016-06-02 | Quick dynamic human body action extraction based on degree of depth study, recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106096518A true CN106096518A (en) | 2016-11-09 |
Family
ID=57447972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610387248.7A Pending CN106096518A (en) | 2016-06-02 | 2016-06-02 | Quick dynamic human body action extraction based on degree of depth study, recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106096518A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107492108A (en) * | 2017-08-18 | 2017-12-19 | 成都通甲优博科技有限责任公司 | A kind of skeleton line extraction algorithm, system and storage medium based on deep learning |
| CN107742296A (en) * | 2017-09-11 | 2018-02-27 | 广东欧珀移动通信有限公司 | Dynamic image generation method and electronic installation |
| CN108655026A (en) * | 2018-05-07 | 2018-10-16 | 上海交通大学 | A kind of quick teaching sorting system of robot and method |
| CN110598569A (en) * | 2019-08-20 | 2019-12-20 | 江西憶源多媒体科技有限公司 | Action recognition method based on human body posture data |
| WO2020062760A1 (en) * | 2018-09-26 | 2020-04-02 | 深圳市中视典数字科技有限公司 | Motion capture system and method |
| CN111496770A (en) * | 2020-04-09 | 2020-08-07 | 上海电机学院 | Intelligent carrying mechanical arm system based on 3D vision and deep learning and use method |
| CN113095675A (en) * | 2021-04-12 | 2021-07-09 | 华东师范大学 | Method for monitoring action mode of examinee by means of identification point in network examination |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101533528A (en) * | 2009-04-18 | 2009-09-16 | 大连大学 | Optical motion capture data processing method based on module piecewise linear model |
| CN101957655A (en) * | 2009-07-17 | 2011-01-26 | 深圳泰山在线科技有限公司 | Marked point-based motion recognition method and terminal equipment |
| US20120057761A1 (en) * | 2010-09-01 | 2012-03-08 | Sony Corporation | Three dimensional human pose recognition method and apparatus |
| CN104463090A (en) * | 2013-11-25 | 2015-03-25 | 安徽寰智信息科技股份有限公司 | Method for recognizing actions of human body skeleton of man-machine interactive system |
| CN105095867A (en) * | 2015-07-21 | 2015-11-25 | 哈尔滨多智科技发展有限公司 | Rapid dynamic face extraction and identification method based deep learning |
-
2016
- 2016-06-02 CN CN201610387248.7A patent/CN106096518A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101533528A (en) * | 2009-04-18 | 2009-09-16 | 大连大学 | Optical motion capture data processing method based on module piecewise linear model |
| CN101957655A (en) * | 2009-07-17 | 2011-01-26 | 深圳泰山在线科技有限公司 | Marked point-based motion recognition method and terminal equipment |
| US20120057761A1 (en) * | 2010-09-01 | 2012-03-08 | Sony Corporation | Three dimensional human pose recognition method and apparatus |
| CN104463090A (en) * | 2013-11-25 | 2015-03-25 | 安徽寰智信息科技股份有限公司 | Method for recognizing actions of human body skeleton of man-machine interactive system |
| CN105095867A (en) * | 2015-07-21 | 2015-11-25 | 哈尔滨多智科技发展有限公司 | Rapid dynamic face extraction and identification method based deep learning |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107492108A (en) * | 2017-08-18 | 2017-12-19 | 成都通甲优博科技有限责任公司 | A kind of skeleton line extraction algorithm, system and storage medium based on deep learning |
| CN107742296A (en) * | 2017-09-11 | 2018-02-27 | 广东欧珀移动通信有限公司 | Dynamic image generation method and electronic installation |
| CN108655026A (en) * | 2018-05-07 | 2018-10-16 | 上海交通大学 | A kind of quick teaching sorting system of robot and method |
| CN108655026B (en) * | 2018-05-07 | 2020-08-14 | 上海交通大学 | Robot rapid teaching sorting system and method |
| WO2020062760A1 (en) * | 2018-09-26 | 2020-04-02 | 深圳市中视典数字科技有限公司 | Motion capture system and method |
| CN110598569A (en) * | 2019-08-20 | 2019-12-20 | 江西憶源多媒体科技有限公司 | Action recognition method based on human body posture data |
| CN110598569B (en) * | 2019-08-20 | 2022-03-08 | 江西憶源多媒体科技有限公司 | Action recognition method based on human body posture data |
| CN111496770A (en) * | 2020-04-09 | 2020-08-07 | 上海电机学院 | Intelligent carrying mechanical arm system based on 3D vision and deep learning and use method |
| CN113095675A (en) * | 2021-04-12 | 2021-07-09 | 华东师范大学 | Method for monitoring action mode of examinee by means of identification point in network examination |
| CN113095675B (en) * | 2021-04-12 | 2022-03-29 | 华东师范大学 | Method for monitoring action mode of examinee by means of identification point in network examination |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106096518A (en) | Quick dynamic human body action extraction based on degree of depth study, recognition methods | |
| Singh et al. | Eye in the sky: Real-time drone surveillance system (dss) for violent individuals identification using scatternet hybrid deep learning network | |
| Bilinski et al. | Human violence recognition and detection in surveillance videos | |
| CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
| Li et al. | Robust visual tracking based on convolutional features with illumination and occlusion handing | |
| CN110517292A (en) | Method for tracking target, device, system and computer readable storage medium | |
| CN106651913A (en) | Target tracking method based on correlation filtering and color histogram statistics and ADAS (Advanced Driving Assistance System) | |
| WO2017129020A1 (en) | Human behaviour recognition method and apparatus in video, and computer storage medium | |
| CN107909059A (en) | It is a kind of towards cooperateing with complicated City scenarios the traffic mark board of bionical vision to detect and recognition methods | |
| Rodriguez et al. | Detecting and segmenting humans in crowded scenes | |
| CN108875586B (en) | Functional limb rehabilitation training detection method based on depth image and skeleton data multi-feature fusion | |
| CN110795982A (en) | Apparent sight estimation method based on human body posture analysis | |
| CN108681700A (en) | A kind of complex behavior recognition methods | |
| CN108280808B (en) | Method for tracking target based on structuring output correlation filter | |
| CN103020606A (en) | Pedestrian detection method based on spatio-temporal context information | |
| CN105303163B (en) | A kind of method and detection device of target detection | |
| CN103226713A (en) | Multi-view behavior recognition method | |
| CN104392445A (en) | Method for dividing crowd in surveillance video into small groups | |
| CN106327528A (en) | Moving object tracking method and operation method of unmanned aerial vehicle | |
| Duan et al. | A more accurate mask detection algorithm based on Nao robot platform and YOLOv7 | |
| CN104616323B (en) | A kind of time and space significance detection method based on slow signature analysis | |
| Nguyen et al. | Facemask wearing alert system based on simple architecture with low-computing devices | |
| CN109389048A (en) | Pedestrian detection and tracking in a kind of monitor video | |
| CN102222324B (en) | Symmetry property-based method for detecting salient regions of images | |
| Yin et al. | YOLO-EPF: Multi-scale smoke detection with enhanced pool former and multiple receptive fields |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 519000 Guangdong city of Zhuhai province Hengqin Baohua Road No. 6, room 105, -23248 (central office) Applicant after: Zhuhai wisdom Technology Co., Ltd. Address before: 150000 Harbin, Nangang District, Thai Garden, the sea floor of the 25 floor, No. 3, No. 1 shops Applicant before: HARBIN DUOZHI SCIENCE AND TECHNOLOGY DEVELOPMENT CO., LTD. |
|
| 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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20161109 |