CN110298214A - A kind of stage multi-target tracking and classification method based on combined depth neural network - Google Patents
A kind of stage multi-target tracking and classification method based on combined depth neural network Download PDFInfo
- Publication number
- CN110298214A CN110298214A CN201810242992.7A CN201810242992A CN110298214A CN 110298214 A CN110298214 A CN 110298214A CN 201810242992 A CN201810242992 A CN 201810242992A CN 110298214 A CN110298214 A CN 110298214A
- Authority
- CN
- China
- Prior art keywords
- neural network
- performer
- stage
- method based
- classification method
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of stage multi-target tracking and classification method based on combined depth neural network, hardware device includes: one or more video camera and calculation server in the method;The method are as follows: in conjunction with a variety of deep neural networks, and use image processing techniques, by the training to deep neural network model, realize that more performers track simultaneously in stage performance and role distinguishes.The present invention solves detection, tracking and the Classification and Identification of the different type performer during stage is had a dress rehearsal and performed in conjunction with computer vision algorithms make by a variety of neural network algorithms.This technology can understand for subsequent stage, scenery and the intelligent stage applications such as illumination scheme intelligently generates, stage choreography intelligently assists provide important basis.
Description
Technical field
The present invention relates to the tracking of performer and Classification and Identification technical field, specially a kind of to be based on combined depth neural network
Stage multi-target tracking and classification method.
Background technique
One excellent stage performance be unable to do without outstanding scenery and light of stage in visual aspects.And perform personnel
To walk be one of the key factor of scenery and light of stage design with role's distribution.Traditional scenery design and dance
Platform Lighting Design often designs suitable setting and light according to the arrangement that walks of performer when having a dress rehearsal by veteran designer
Effect.However, the design of this immobilization, the performance of performer is restricted in performance very much, performer needs according to dress rehearsal
It walks and is performed.Also often occur performer in some large-scale performances and does not go to the feelings on correct position and missing light
Condition.
Currently, for this problem, the method for mainstream is being waved by wearing the realization performer of short-distance wireless alignment sensor
Positioning on platform.But the electromagnetic environment interference for being highly susceptible to stage complexity of this method, and be unfavorable for for more performers
Scene use, can not also identify the role of performer.
For the present invention by the way that the deep neural network algorithms of different purposes is applied in combination, solve stage performer is positioned at angle
Color identifies problem.Important information foundation is provided for subsequent stage intelligent.
Summary of the invention
The purpose of the present invention is to provide a kind of stage multi-target trackings based on combined depth neural network and classification side
Method, to solve the problems mentioned in the above background technology.Information foundation is provided for subsequent intelligent stage application.
To achieve the above object, the invention provides the following technical scheme:
A kind of stage multi-target tracking and classification method based on combined depth neural network, hardware device in the method
It include: one or more video camera and calculation server;The method are as follows: in conjunction with a variety of deep neural networks, and use
Image processing techniques realizes more performers while tracking and role in stage performance by the training to deep neural network model
It distinguishes.
As a further solution of the present invention: specifically comprising the following steps:
1) data acquire: shooting object scene by video camera, image includes that institute's some need is tracked and what is distinguished drills
Member;
2) performer's position detection and image segmentation: by the target detection deep neural network of reasonable training, every frame is found
The position of all performers in image, and intercept the picture of each performer;
3) picture after interception segmentation: being carried out the pretreatment of uniform sizes by image preprocessing after dividing, and is guaranteeing personage
In the case where ratio, image scaling is carried out;
4) performer's position tracking: by the target following neural network of reasonable training, the movement rail of each performer is obtained
Mark
5) performer role distinguishes: carrying out Unsupervised clustering analysis to performer role, analyzes different performer's types.
As a further solution of the present invention: the performer in the method includes singer, dancer or actor.
As a further solution of the present invention: the video camera uses high definition high-speed camera.
As a further solution of the present invention: target detection deep neural network uses Fast-RCNN in the step 2)
Neural network, Faster-RCNN neural network or YOLO neural network.
As a further solution of the present invention: the training data in the step 2) is from various large-scale concerts, performance
A large amount of video and image are intercepted, is marked by human assistance, training sample abundant is formed.
As a further solution of the present invention: target following neural network, which uses, in the step 4) is based on correlation filter
End-to-end track algorithm, or using the track algorithm based on trained and high speed depth Recurrent networks.
Compared with prior art, the beneficial effects of the present invention are:
(1) it is tracked and Classification and Identification using contactless method;
(2) real-time tracking of multiple target performer is realized;
(3) the accurate differentiation of performer role is realized.
Detailed description of the invention
Fig. 1 is the schematic diagram of the method for the present invention usage scenario;
Fig. 2 is the corresponding system flow chart of the method for the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Please refer to Fig. 1~2, the present invention provides a kind of technical solution: a kind of stage based on combined depth neural network is more
Target tracking and classification method, hardware device includes: one or more video camera and calculation server in the method;It is described
Method are as follows: in conjunction with a variety of deep neural networks, and use image processing techniques, pass through the instruction to deep neural network model
Practice, realizes that more performers track simultaneously in stage performance and role distinguishes.
Specific implementation includes the following steps:
1) data acquire: shooting object scene by video camera, image includes that institute's some need is tracked and what is distinguished drills
Member;
2) performer's position detection and image segmentation: by the target detection deep neural network of reasonable training, every frame is found
The position of all performers in image, and intercept the picture of each performer;
3) picture after interception segmentation: being carried out the pretreatment of uniform sizes by image preprocessing after dividing, and is guaranteeing personage
In the case where ratio, image scaling is carried out;
4) performer's position tracking: by the target following neural network of reasonable training, the movement rail of each performer is obtained
Mark
5) performer role distinguishes: carrying out Unsupervised clustering analysis to performer role, analyzes different performer's types.
Above-described " performer " includes the various performers such as singer, dancer, actor.
In the step 1), to the use of the model, quantity of video camera there is no limit, it is only necessary to cover the whole audience, most
High definition high-speed camera is used well, can improve the precision of identification.
In the step 2), the type for the target detection neural network that can be used has: Fast-RCNN network (reference:
Girshick R.Fast R-CNN[C]//Computer Vision(ICCV),2015IEEE International
Conference on.IEEE, 2015:1440-1448.), (reference: Ren S, He K, Girshick of Faster-RCNN network
R,et al.Faster R-CNN:towards real-time object detection with region proposal
networks[J].IEEE transactions on pattern analysis and machine intelligence,
2017,39 (6): 1137-1149.), (reference: Redmon J, Divvala S, Girshick R, et al.You of YOLO network
only look once:Unified,real-time object detection[C]//Proceedings of the IEEE
Conference on computer vision and pattern recognition.2016:779-788.) etc. nerve nets
Network method.
In the step 2), training data can intercept a large amount of video and image from various large-scale concerts, performance,
It is marked by human assistance, forms training sample abundant.
In the step 4), the monotrack algorithm of various maturations can be used, such as: the end based on correlation filter
To end track algorithm (Valmadre J, Bertinetto L, Henriques J, et al.End-to-end
representation learning for correlation filter based tracking[C]//Computer
Vision and Pattern Recognition(CVPR),2017IEEE Conference on.IEEE,2017:5000-
5008.), based on trained and high speed depth Recurrent networks track algorithm (Held D, Thrun S, Savarese
S.Learning to track at 100fps with deep regression networks[C]//European
Conference on Computer Vision.Springer,Cham,2016:749-765.)
The important difference of the step 4) and step 2) is: step 2) is only to have found position, but not can determine that
Whether whether certain performers have carried out location swap, and step 4) meeting track individual feature, the movement for carrying out single performer chase after
Track, even if can also accurately be tracked under similar action situation in similar clothing.
In the step 5), unsupervised clustering algorithm can be used to distinguish role different on stage.The area of role
Following several classes: star and accompanying dancer, protagonist and supporting role, combination team etc. can be divided into.In order to be not limited to fixed classification mould
Formula, therefore use Unsupervised clustering algorithm.Such as use classical K-means algorithm (Ahmad A, Dey L.A k-mean
clustering algorithm for mixed numeric and categorical data[J].Data&Knowledge
Engineering,2007,63(2):503-527.)。
It may finally be the motion profile for analyzing every performer in real time in stage performance by above scheme.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.
Claims (7)
1. a kind of stage multi-target tracking and classification method based on combined depth neural network, it is characterised in that: the method
Middle hardware device includes: one or more video camera and calculation server;The method are as follows: in conjunction with a variety of depth nerve nets
Network, and image processing techniques is used, by the training to deep neural network model, realize that more performers are simultaneously in stage performance
Tracking and role distinguish.
2. a kind of stage multi-target tracking and classification method based on combined depth neural network according to claim 1,
It is characterized by: specifically comprising the following steps:
1) data acquire: shooting object scene by video camera, image includes the performer that institute's some need is tracked and distinguishes;
2) performer's position detection and image segmentation: by the target detection deep neural network of reasonable training, every frame image is found
In all performers position, and intercept the picture of each performer;
3) picture after interception segmentation: being carried out the pretreatment of uniform sizes by image preprocessing after dividing, and is guaranteeing personage's ratio
In the case where, carry out image scaling;
4) performer's position tracking: by the target following neural network of reasonable training, the motion profile of each performer is obtained
5) performer role distinguishes: carrying out Unsupervised clustering analysis to performer role, analyzes different performer's types.
3. a kind of stage multi-target tracking and classification method based on combined depth neural network according to claim 2,
It is characterized by: the performer in the method includes singer, dancer or actor.
4. a kind of stage multi-target tracking and classification method based on combined depth neural network according to claim 2,
It is characterized by: the video camera uses high definition high-speed camera.
5. a kind of stage multi-target tracking and classification method based on combined depth neural network according to claim 2,
It is characterized by: target detection deep neural network uses Fast-RCNN neural network, Faster-RCNN in the step 2)
Neural network or YOLO neural network.
6. a kind of stage multi-target tracking and classification method based on combined depth neural network according to claim 2,
It is characterized by: the training data in the step 2) intercepts a large amount of video and image from various large-scale concerts, performance,
It is marked by human assistance, forms training sample abundant.
7. a kind of stage multi-target tracking and classification method based on combined depth neural network according to claim 2,
It is characterized by: target following neural network uses the end-to-end track algorithm based on correlation filter in the step 4), or
Person uses the track algorithm based on trained and high speed depth Recurrent networks.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810242992.7A CN110298214A (en) | 2018-03-23 | 2018-03-23 | A kind of stage multi-target tracking and classification method based on combined depth neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810242992.7A CN110298214A (en) | 2018-03-23 | 2018-03-23 | A kind of stage multi-target tracking and classification method based on combined depth neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110298214A true CN110298214A (en) | 2019-10-01 |
Family
ID=68025806
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810242992.7A Pending CN110298214A (en) | 2018-03-23 | 2018-03-23 | A kind of stage multi-target tracking and classification method based on combined depth neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110298214A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111339687A (en) * | 2020-03-26 | 2020-06-26 | 北京理工大学 | Crowd performance site sparing system based on deep learning |
CN113920164A (en) * | 2021-10-27 | 2022-01-11 | 浙江工商大学 | Actor identity re-identification method based on near-infrared anti-counterfeiting ink in theater environment |
CN113993250A (en) * | 2021-12-24 | 2022-01-28 | 深圳市奥新科技有限公司 | Stage lighting control method, device, equipment and storage medium |
CN114071842A (en) * | 2021-11-23 | 2022-02-18 | 横店集团得邦照明股份有限公司 | Intelligent stage lighting system based on target detection and tracking |
CN114245543A (en) * | 2021-12-27 | 2022-03-25 | 江苏领焰智能科技股份有限公司 | Stage lighting method, system, equipment and storage medium based on space positioning |
CN115881001A (en) * | 2022-12-30 | 2023-03-31 | 深圳市联诚发科技股份有限公司 | Display control method and device of polymorphic display screen and polymorphic display screen |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103489199A (en) * | 2012-06-13 | 2014-01-01 | 通号通信信息集团有限公司 | Video image target tracking processing method and system |
CN103839049A (en) * | 2014-02-26 | 2014-06-04 | 中国计量学院 | Double-person interactive behavior recognizing and active role determining method |
US20160239982A1 (en) * | 2014-08-22 | 2016-08-18 | Zhejiang Shenghui Lighting Co., Ltd | High-speed automatic multi-object tracking method and system with kernelized correlation filters |
CN106097391A (en) * | 2016-06-13 | 2016-11-09 | 浙江工商大学 | A kind of multi-object tracking method identifying auxiliary based on deep neural network |
CN106407891A (en) * | 2016-08-26 | 2017-02-15 | 东方网力科技股份有限公司 | Target matching method based on convolutional neural network and device |
CN106659940A (en) * | 2014-05-21 | 2017-05-10 | 环球城市电影有限责任公司 | Optical tracking system for automation of amusement park elements |
-
2018
- 2018-03-23 CN CN201810242992.7A patent/CN110298214A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103489199A (en) * | 2012-06-13 | 2014-01-01 | 通号通信信息集团有限公司 | Video image target tracking processing method and system |
CN103839049A (en) * | 2014-02-26 | 2014-06-04 | 中国计量学院 | Double-person interactive behavior recognizing and active role determining method |
CN106659940A (en) * | 2014-05-21 | 2017-05-10 | 环球城市电影有限责任公司 | Optical tracking system for automation of amusement park elements |
US20160239982A1 (en) * | 2014-08-22 | 2016-08-18 | Zhejiang Shenghui Lighting Co., Ltd | High-speed automatic multi-object tracking method and system with kernelized correlation filters |
CN106097391A (en) * | 2016-06-13 | 2016-11-09 | 浙江工商大学 | A kind of multi-object tracking method identifying auxiliary based on deep neural network |
CN106407891A (en) * | 2016-08-26 | 2017-02-15 | 东方网力科技股份有限公司 | Target matching method based on convolutional neural network and device |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111339687A (en) * | 2020-03-26 | 2020-06-26 | 北京理工大学 | Crowd performance site sparing system based on deep learning |
CN113920164A (en) * | 2021-10-27 | 2022-01-11 | 浙江工商大学 | Actor identity re-identification method based on near-infrared anti-counterfeiting ink in theater environment |
CN113920164B (en) * | 2021-10-27 | 2024-05-24 | 浙江工商大学 | Actor identity re-identification method based on near infrared anti-counterfeiting ink in theatre environment |
CN114071842A (en) * | 2021-11-23 | 2022-02-18 | 横店集团得邦照明股份有限公司 | Intelligent stage lighting system based on target detection and tracking |
CN113993250A (en) * | 2021-12-24 | 2022-01-28 | 深圳市奥新科技有限公司 | Stage lighting control method, device, equipment and storage medium |
CN113993250B (en) * | 2021-12-24 | 2022-03-15 | 深圳市奥新科技有限公司 | Stage lighting control method, device, equipment and storage medium |
CN114245543A (en) * | 2021-12-27 | 2022-03-25 | 江苏领焰智能科技股份有限公司 | Stage lighting method, system, equipment and storage medium based on space positioning |
CN115881001A (en) * | 2022-12-30 | 2023-03-31 | 深圳市联诚发科技股份有限公司 | Display control method and device of polymorphic display screen and polymorphic display screen |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110298214A (en) | A kind of stage multi-target tracking and classification method based on combined depth neural network | |
Manafifard et al. | A survey on player tracking in soccer videos | |
Zhao et al. | Crossing-line crowd counting with two-phase deep neural networks | |
CN100390811C (en) | Method for tracking multiple human faces from video in real time | |
Senst et al. | Detecting people carrying objects based on an optical flow motion model | |
Vats et al. | Player tracking and identification in ice hockey | |
CN103761514A (en) | System and method for achieving face recognition based on wide-angle gun camera and multiple dome cameras | |
Zhang et al. | Robust multi-object tracking via cross-domain contextual information for sports video analysis | |
Lian et al. | Spatial–temporal consistent labeling of tracked pedestrians across non-overlapping camera views | |
CN109325456A (en) | Target identification method, device, target identification equipment and storage medium | |
CN109344842A (en) | A kind of pedestrian's recognition methods again based on semantic region expression | |
Liu et al. | Soccer video event detection using 3D convolutional networks and shot boundary detection via deep feature distance | |
Shah et al. | Multi-view action recognition using contrastive learning | |
Liang et al. | Methods of moving target detection and behavior recognition in intelligent vision monitoring. | |
Zhou et al. | A survey of multi-object video tracking algorithms | |
CN109753962A (en) | Text filed processing method in natural scene image based on hybrid network | |
Wang | Exploring intelligent image recognition technology of football robot using omnidirectional vision of internet of things | |
Prasanna et al. | RETRACTED ARTICLE: An effiecient human tracking system using Haar-like and hog feature extraction | |
CN108830222A (en) | A kind of micro- expression recognition method based on informedness and representative Active Learning | |
Mu et al. | Resgait: The real-scene gait dataset | |
Cheng et al. | Agricultural pests tracking and identification in video surveillance based on deep learning | |
CN108765459A (en) | Semi-online visual multi-target tracking method based on small trajectory graph association model | |
Praveenkumar et al. | Real-time multi-object tracking of pedestrians in a video using convolution neural network and Deep SORT | |
CN108491832A (en) | A kind of embedded human face identification follow-up mechanism and method | |
Yang et al. | Design and implementation of intelligent analysis technology in sports video target and trajectory tracking algorithm |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191001 |