CN107016687A - The hybrid algorithm of video frequency motion target detect and track - Google Patents
The hybrid algorithm of video frequency motion target detect and track Download PDFInfo
- Publication number
- CN107016687A CN107016687A CN201710216663.0A CN201710216663A CN107016687A CN 107016687 A CN107016687 A CN 107016687A CN 201710216663 A CN201710216663 A CN 201710216663A CN 107016687 A CN107016687 A CN 107016687A
- Authority
- CN
- China
- Prior art keywords
- algorithm
- track
- image
- video frequency
- color
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of hybrid algorithm of video frequency motion target detect and track.The hybrid algorithm of the video frequency motion target detect and track, which is integrated, uses mean shift algorithm and color histogram nomography, stability and precision that characteristic matching keeps tracking are carried out with mean shift algorithm, ensures the distribution of color and dynamic video of target and the rotation of image the adaptability of viewing angle with color histogram nomography.Compared with correlation technique, the hybrid algorithm for the video frequency motion target detect and track that the present invention is provided, which combines mean shift algorithm, can keep the precision and stability and color histogram of tracking the change of rotation and the observation of dynamic video image is had the advantages that good adaptability, and more effectively moving target is tracked.
Description
Technical field
Calculated the present invention relates to the mixing in Virtual Environment field, more particularly to a kind of video frequency motion target detect and track
Method.
Background technology
With the development of athletic competition level, the requirement to training is also improved constantly.The training of past coach is
The need for it can not meet training.Computer vision technique is more accurate than human eye, remembers more longlasting, it can faster, more have
Effect ground catches mobile object.By record the various motions of target data for the motion of sportsman provide it is more theoretical and
Describe to adapt to current form based on data.
Therefore, it is necessary to provide a kind of hybrid algorithm of video frequency motion target detect and track to meet above demand.
The content of the invention
It is an object of the invention to provide a kind of video motion based on mean shift algorithm and color histogram nomography
Target detection and the hybrid algorithm of tracking.
The present invention provides a kind of hybrid algorithm of video frequency motion target detect and track, it is characterised in that comprehensive using equal
Value drift algorithm and color histogram nomography, stability and essence that characteristic matching keeps tracking are carried out with mean shift algorithm
Degree, ensures the distribution of color and dynamic video of target and the rotation of image the adaptation of viewing angle with color histogram nomography
Property, wherein, the mean shift algorithm includes：
Step 1.1, valuation f of the polynary kernel function density in some point x can be obtained according to core K (x) and windows radius H
(x), its calculation formula isFirst calculate the skew average of current point；
Step 1.2, then move the point to its skew average；
Step 1.3, the compiling average is new starting point, is continued to move to；
Step 1.4, until the end that meets some requirements；
The color histogram nomography includes：
Step 2.1, average global error minimum value between density estimation and real density is calculated, its calculation formula is：
Step 2.2, white value is calculated, i.e. D ties up the volume of spheroid, represented with the core of many Gausses, formula is：The profile function of Gaussian kernel is expressed as：
Step 2.3, the gradient of probability density is defined using kernel function density gradient valuation：Formula is
Wherein
Compared with correlation technique, the hybrid algorithm for the video frequency motion target detect and track that the present invention is provided is to be based on average
The mixing of migration algorithm and color histogram nomography, mean shift algorithm can keep the precision and stability of tracking, and color is straight
Square figure makes the change of rotation and the observation of dynamic video image have good adaptability, and both combine, more effectively to motion
Target is tracked.
Brief description of the drawings
The schematic diagram of Image Engineering in the hybrid algorithm for the video frequency motion target detect and track that Fig. 1 provides for the present invention；
Image Compression structure in the hybrid algorithm for the video frequency motion target detect and track that Fig. 2 provides for the present invention
Figure；
Image processing process principle in the hybrid algorithm for the video frequency motion target detect and track that Fig. 3 provides for the present invention
Figure；
The hybrid algorithm for the video frequency motion target detect and track that Fig. 4 provides for the present invention is to portable light display device
Experimental result；
The progress on MATLAB platforms of the hybrid algorithm for the video frequency motion target detect and track that Fig. 5 provides for the present invention
A kind of video tracking emulation experiment；
The progress on MATLAB platforms of the hybrid algorithm for the video frequency motion target detect and track that Fig. 6 provides for the present invention
Another video tracking emulation experiment.
Embodiment
Describe the present invention in detail below with reference to accompanying drawing and in conjunction with the embodiments.It should be noted that in the feelings not conflicted
Under condition, the embodiment in the present invention and the feature in embodiment can be mutually combined.For sake of convenience, hereinafter as occurred
" on ", " under ", "left", "right" printed words, only represent that the upper and lower, left and right direction with accompanying drawing in itself is consistent, does not limit structure
It is set for using.
The hybrid algorithm for the video frequency motion target detect and track that the present invention is provided, it is comprehensive to use mean shift algorithm and face
Color Histogram algorithm, carries out stability and precision that characteristic matching keeps tracking, with color histogram with mean shift algorithm
Nomography ensures the distribution of color and dynamic video of target and the rotation of image the adaptability of viewing angle.
Mean shift algorithm is a kind of nonparametric technique risen based on density gradient, and probability is calculated by interative computation
The extreme point of density function finds target location, realizes target following.The algorithm has the spy of printenv, Fast Pattern Matching
Point, has a wide range of applications in the field such as computer vision and Computer Image Processing.Mean shift algorithm can be along shortest path
The probability distribution in footpath is moved to each point the local maximum point of density function.Our usually said mean shift algorithms are one
Individual iterative step, i.e., first calculate the skew average of current point, the mobile point offsets average to it, then as new starting
Point, is continued to move to, until the end that meets some requirements.
Wherein, the mean shift algorithm includes：
Step 1.1, valuation f of the polynary kernel function density in some point x can be obtained according to core K (x) and windows radius H
(x), its calculation formula isFirst calculate the skew average of current point；
Step 1.2, then move the point to its skew average；
Step 1.3, the compiling average is new starting point, is continued to move to；
Step 1.4, until the end that meets some requirements；
The color histogram nomography includes：
Step 2.1, average global error minimum value between density estimation and real density is calculated, its calculation formula is：
Step 2.2, white value is calculated, i.e. D ties up the volume of spheroid, represented with the core of many Gausses, formula is：The profile function of Gaussian kernel is expressed as：
Step 2.3, the gradient of probability density is defined using kernel function density gradient valuation：Formula is
Wherein
While using mean shift algorithm and color histogram nomography, Image Engineering calculating is also carried out, is specifically included
Color difference image is represented, compression of images and processing.
Refer to Image Engineering principle in the hybrid algorithm for the video frequency motion target detect and track that Fig. 1 provides for the present invention
Figure.Image Engineering is divided into three levels according to abstract and research method：Image procossing, graphical analysis and image understanding.In reality
In, the competition and training video recording of sportsman in training process can be obtained.Video camera is erect with identical height and angle
In competition area or training court, by image procossing, all kinds of reference datas are obtained., can according to the height and angle as pole
To represent the specific coordinate of sports video object space with data relationship.Researcher can pass through computer vision technique, number
Learn image procossing and mode identification technology to handle the graphical analysis collected, so as to obtain required statistics.
Moving object detection and tracking system are based on technologies such as Digital Image Processing, pattern-recognition, computer visions
Intelligent identifying system.The system can be widely applied to traffic control, astronomical observation, biomedical research, traffic statistics and body
The association area such as educate.In sports video analysis field, moving object detection and tracking technique serve indispensable effect.It is logical
The real-time detect and track to Athletess is crossed, the movement locus of sportsman can be analyzed, facilitate the instruction of sportsman
Practice.Human eye can not detect action difference trickle during the games, it is necessary to improve constantly the training and competition effect of sportsman
Really.Therefore, the object detecting and tracking technology that the present invention is provided can bring larger practical value.
The brightness of each location of pixels is unrelated with modulation function in image, and the weight coefficient estimation of brightness is uniform.
On the basis of this, the convolution kernel of 3x3 normalization space symmetr can be used to handle, the brightness of each location of pixels can be with
Calculated, color difference image is expressed as：
ΔCFA(x, y)=∑ ΔCFA(x, y) fS(x, y)=cRfR(x, y)+cBfB(x, y)+cGfG(x, y)
Wherein, CR, CG and CB are the color difference image of modulation：
CR, CG and CB, calculation formula be respectively：
From CR, CG and CB, calculation formula can release Bel's mould after sampling by adjusting color difference image formation
Plate color difference image：
Compression of images is a kind of technology for referring to and original picture element matrix being damaged or nondestructively represented with less bit, also referred to as
Image Coding.The data volume needed when representing digital picture the purpose is to reduce.General digital picture redundancy has spatial domain
Redundancy and frequency domain redundancy, with temporal correlation.Compress technique is exactly that the redundancy removed in data (removes divisor
Correlation between), to reduce data rate.Compression of images can be divided into damage data compression and lossless data compression, wherein compiling
The image compression algorithm that code and the distortion level of decoding are consistent can all be divided into Lossless Compression.The compression process of Lossless Compression
It is reversible, and the loss of information then occurs in lossy compression method.Modern coding is encoded to from classics, the efficiency of compress technique has
It is many to improve.
Refer to Image Compression in the hybrid algorithm for the video frequency motion target detect and track that Fig. 2 provides for the present invention
Structure chart.In the present embodiment, compression of images uses Lossless Compression, and also known as reversible pressure compression, is data volume needed for a kind of reduce
Method, represent required image data amount using the lossless image of low bit rate.The working mechanism of Lossless Compression is to eliminate number
According to redundancy.In compression process, due to only using redundancy between coding redundancy and pixel, any loss is not caused, therefore can be with
Undistorted decoding is carried out to image, recovers raw image data.In practice, Image Compression develops from simple entropy code
To transition coding, predictive coding, quantization encoding, multi-resolution encoding.
Refer to image processing process in the hybrid algorithm for the video frequency motion target detect and track that Fig. 3 provides for the present invention
Schematic diagram.Image segmentation is from image preprocessing to image recognition and the committed step of graphical analysis, is in image procossing
Very important status.On the one hand, it is the basis of expression, and pattern measurement is had a major impact.On the other hand, it makes based on figure
Objective expression, feature extraction and parameter testing of picture etc. by original image be converted into more abstract, greater compactness of form turn into can
Can, also make it possible the image recognition and graphical analysis of higher level.
The hybrid algorithm for referring to the video frequency motion target detect and track that Fig. 4 provides for the present invention is aobvious to portable light
Show equipment experimental result.Because the motion of human body is the target of a non-rigid, as can be seen from Figure 1 target shift position
Suddenly change, shape is also in change, so tracking is difficult.From tracking result as can be seen that the video motion that the present invention is provided
Target detection and the hybrid algorithm of tracking can accurately track the position of target, and be tracked according to the Adjusting Shape of people in present frame
Window.As can be seen that the algorithm can adapt to different goal-settings, so as to demonstrate the calculation from Basketball Match video tracking
Method has preferable application effect.
It is the hybrid algorithm of video frequency motion target detect and track that the present invention is provided in MATLAB platforms to refer to Fig. 5
A kind of video tracking emulation experiment of upper progress.The hybrid algorithm that the present invention is provided uses mean shift algorithm and color histogram
Algorithm, the feature of moving object tracking.Mean shift algorithm Feature Correspondence Algorithm and track algorithm, can keep the stabilization of tracking
Property and precision.Distribution of color and dynamic video and the rotation of image of the color histogram nomography mainly for target, become to angle
The observation of change has good adaptability.MATLAB simulation accuracies demonstrate the motion that bar is substantially increased after two kinds of algorithms are combined
Object tracking.Test result indicates that, this method has preferable effect.
It is the hybrid algorithm of video frequency motion target detect and track that the present invention is provided in MATLAB platforms to refer to Fig. 6
Another video tracking emulation experiment of upper progress.The complexity of physical culture in itself is come for the actually detected and track band of mobile object
Many difficulties.For effectively detect and track sportsman, present invention improves over common single track algorithm, it is proposed that one
The tracking for combining mean shift algorithm and color histogram nomography is planted, the detect and track effect of Moving Objects is improved.
Embodiments of the invention are the foregoing is only, are not intended to limit the scope of the invention, it is every to utilize this hair
Equivalent structure or equivalent flow conversion that bright specification and accompanying drawing content are made, or directly or indirectly it is used in other related skills
Art field, is included within the scope of the present invention.
Claims (3)
1. a kind of hybrid algorithm of video frequency motion target detect and track, it is characterised in that calculated including comprehensive using average drifting
Method and color histogram nomography, carry out stability and precision that characteristic matching keeps tracking, with face with mean shift algorithm
Color Histogram algorithm ensures the distribution of color and dynamic video of target and the rotation of image the adaptability of viewing angle, wherein,
The mean shift algorithm includes：
Step 1.1, valuation f (x) of the polynary kernel function density in some point x can be obtained according to core K (x) and windows radius H, its
Calculation formula isFirst calculate the skew average of current point；
Step 1.2, then move the point to its skew average；
Step 1.3, the compiling average is new starting point, is continued to move to；
Step 1.4, until the end that meets some requirements；
The color histogram nomography includes：
Step 2.1, average global error minimum value between density estimation and real density is calculated, its calculation formula is：
Step 2.2, white value is calculated, i.e. D ties up the volume of spheroid, represented with the core of many Gausses, formula is：The profile function of Gaussian kernel is expressed as：
Step 2.3, the gradient of probability density is defined using kernel function density gradient valuation：Formula is
Wherein
2. the hybrid algorithm of video frequency motion target detect and track according to claim 1, it is characterised in that using equal
While value drift algorithm and color histogram nomography, also carry out Image Engineering calculating, specifically include color difference image represent, image
Compression and processing.
3. the hybrid algorithm of video frequency motion target detect and track according to claim 2, it is characterised in that color difference image
It is expressed as：
ΔCFA(x, y)=∑ ΔCFA(x, y) fS(x, y)=cRfR(x, y)+cBfB(x, y)+cGfG(x, y)
Wherein, CR, CG and CB are the color difference image of modulation：
CR, CG and CB, calculation formula be respectively：
From CR, CG and CB, calculation formula can release a Bayer pattern color after sampling by adjusting color difference image formation
Difference image：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710216663.0A CN107016687A (en) | 2017-03-25 | 2017-03-25 | The hybrid algorithm of video frequency motion target detect and track |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710216663.0A CN107016687A (en) | 2017-03-25 | 2017-03-25 | The hybrid algorithm of video frequency motion target detect and track |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107016687A true CN107016687A (en) | 2017-08-04 |
Family
ID=59445827
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710216663.0A Pending CN107016687A (en) | 2017-03-25 | 2017-03-25 | The hybrid algorithm of video frequency motion target detect and track |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107016687A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110807790A (en) * | 2019-10-31 | 2020-02-18 | 智慧视通(杭州)科技发展有限公司 | Image data extraction and compression method for video target trajectory tracking content |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101577006A (en) * | 2009-06-15 | 2009-11-11 | 北京中星微电子有限公司 | Loitering detecting method and loitering detecting system in video monitoring |
| CN102289822A (en) * | 2011-09-09 | 2011-12-21 | 南京大学 | Method for tracking moving target collaboratively by multiple cameras |
| CN102629385A (en) * | 2012-02-28 | 2012-08-08 | 中山大学 | Object matching and tracking system based on multiple camera information fusion and method thereof |
| CN103426179A (en) * | 2012-05-17 | 2013-12-04 | 深圳中兴力维技术有限公司 | Target tracking method and system based on mean shift multi-feature fusion |
| CN104166996A (en) * | 2014-08-06 | 2014-11-26 | 北京航空航天大学 | Human eye tracking method based on edge and color double-feature space column diagram |
| CN106488117A (en) * | 2015-08-27 | 2017-03-08 | 安讯士有限公司 | The preprocess method of digital picture and digital Image semantic classification system |
-
2017
- 2017-03-25 CN CN201710216663.0A patent/CN107016687A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101577006A (en) * | 2009-06-15 | 2009-11-11 | 北京中星微电子有限公司 | Loitering detecting method and loitering detecting system in video monitoring |
| CN102289822A (en) * | 2011-09-09 | 2011-12-21 | 南京大学 | Method for tracking moving target collaboratively by multiple cameras |
| CN102629385A (en) * | 2012-02-28 | 2012-08-08 | 中山大学 | Object matching and tracking system based on multiple camera information fusion and method thereof |
| CN103426179A (en) * | 2012-05-17 | 2013-12-04 | 深圳中兴力维技术有限公司 | Target tracking method and system based on mean shift multi-feature fusion |
| CN104166996A (en) * | 2014-08-06 | 2014-11-26 | 北京航空航天大学 | Human eye tracking method based on edge and color double-feature space column diagram |
| CN106488117A (en) * | 2015-08-27 | 2017-03-08 | 安讯士有限公司 | The preprocess method of digital picture and digital Image semantic classification system |
Non-Patent Citations (2)
| Title |
|---|
| 刘国军: "基于可移动拍摄大场景下的人体运动跟踪关键技术的研究与应用", 《中国博士学位论文全文数据库信息科技辑》 * |
| 韩明: "基于水平集表示和均值漂移的运动目标检测与跟踪研究", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110807790A (en) * | 2019-10-31 | 2020-02-18 | 智慧视通(杭州)科技发展有限公司 | Image data extraction and compression method for video target trajectory tracking content |
| CN110807790B (en) * | 2019-10-31 | 2022-06-03 | 智慧视通(杭州)科技发展有限公司 | Image data extraction and compression method for video target trajectory tracking content |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021129064A9 (en) | Posture acquisition method and device, and key point coordinate positioning model training method and device | |
| CN106780620B (en) | Table tennis motion trail identification, positioning and tracking system and method | |
| CN106997597B (en) | It is a kind of based on have supervision conspicuousness detection method for tracking target | |
| CN105869178B (en) | A kind of complex target dynamic scene non-formaldehyde finishing method based on the convex optimization of Multiscale combination feature | |
| CN103514441B (en) | Facial feature point locating tracking method based on mobile platform | |
| Chen et al. | Sportscap: Monocular 3d human motion capture and fine-grained understanding in challenging sports videos | |
| WO2016034059A1 (en) | Target object tracking method based on color-structure features | |
| Birchfield et al. | Spatial histograms for region‐based tracking | |
| CN107622229A (en) | A kind of video frequency vehicle based on fusion feature recognition methods and system again | |
| CN107316005B (en) | Behavior identification method based on dense track kernel covariance descriptor | |
| Nie et al. | A robust and efficient framework for sports-field registration | |
| CN104794737A (en) | Depth-information-aided particle filter tracking method | |
| CN105279769A (en) | Hierarchical particle filtering tracking method combined with multiple features | |
| Zhao et al. | Generative tracking of 3D human motion by hierarchical annealed genetic algorithm | |
| Li et al. | RGBD relocalisation using pairwise geometry and concise key point sets | |
| CN106127766B (en) | Method for tracking target based on Space Coupling relationship and historical models | |
| Hsu et al. | Online view-invariant human action recognition using rgb-d spatio-temporal matrix | |
| CN102156991A (en) | Quaternion based object optical flow tracking method | |
| Potje et al. | Extracting deformation-aware local features by learning to deform | |
| Zheng et al. | Self-supervised monocular depth estimation based on combining convolution and multilayer perceptron | |
| CN107016687A (en) | The hybrid algorithm of video frequency motion target detect and track | |
| CN110910489B (en) | Monocular vision-based intelligent court sports information acquisition system and method | |
| CN108573217B (en) | Compression tracking method combined with local structured information | |
| CN103345762B (en) | Bayes's visual tracking method based on manifold learning | |
| Meng et al. | Analysis of basketball technical movements based on human-computer interaction with deep learning |
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: 20170804 |