CN106504269B - A kind of method for tracking target of more algorithms cooperation based on image classification - Google Patents
A kind of method for tracking target of more algorithms cooperation based on image classification Download PDFInfo
- Publication number
- CN106504269B CN106504269B CN201610917111.8A CN201610917111A CN106504269B CN 106504269 B CN106504269 B CN 106504269B CN 201610917111 A CN201610917111 A CN 201610917111A CN 106504269 B CN106504269 B CN 106504269B
- Authority
- CN
- China
- Prior art keywords
- target
- image
- variation
- tracking
- classification
- 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.)
- Active
Links
Classifications
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20224—Image subtraction
Landscapes
- Image Analysis (AREA)
Abstract
A kind of method for tracking target of more algorithms cooperation based on image classification, comprising steps of (1) initializes all target components;(2) judge whether current goal is front cross frame: if so, using BW-SOAMS algorithm keeps track target;If it is not, the image i.e. after the second frame of input, then classify to image sequence judgement;(3) corresponding Moving Target Tracking Algorithm is called according to classification;(4) real-time update of target's feature-extraction, sample and object candidate area;(5) position of tracking target is provided, next frame is read in, continues through b judgement classification tracking until the last frame of video sequence.
Description
Technical field
The present invention relates to target tracking domains, in particular to a kind of target of more algorithms cooperation based on image classification
Tracking.
Background technique
Target following is one of key problem of computer vision as a cross-cutting cutting edge technology interdisciplinary.Mesh
Mark tracking technique develops since the sixties in last century, till now, has formd a series of method.Due to being tracked mesh
The diversity of sample body variation and the complexity of external environment, motion target tracking is the project for being rich in challenge.One
The target tracking algorism of a robust has to preferably solve all difficulties (such as rotationally-varying, ruler that tracking process encounters
Very little variation, illumination variation, block variation, context similarity variation etc.).Classical target tracking algorism mainly has average drifting calculation
The methods of method, optical flow method, Kalman filtering method and sparse matrix expression, every kind of classical Moving Target Tracking Algorithm has certainly
Oneself advantage and disadvantage, and present Moving Target Tracking Algorithm is essentially all to improve either hair based on above-mentioned classic algorithm
Exhibition.Although in target tracking domain, there are a large amount of target tracking algorisms, the fortune being also able to achieve under certain complex environment
The tracking of moving-target.But the good wide target tracking algorism of adaptation environment of a robust performance is still computer vision
The research topic of one great challenge.The present invention is handled sequence of video images from the source of tracking image sequence,
With classifying in advance to the target and background to be tracked before target following according to the difference because playing image change reason.
Because the reason of playing environmental change can be mainly divided into two classes in object tracking process, that is, track target Self-variation
With the variation of tracking environmental, in actual test we have found that because the factor for playing target Self-variation mainly has tracking target
Variation and change in size are blocked, because playing mainly being changed by illumination variation and context similarity for environmental change etc..Become according to two kinds
Change and image is not divided into two classes before tracking with us because of image change range: image overall variation and part are interested
Regional change.Person's appropriate algorithm is not then selected to carry out tracking and positioning to video sequence, in this way for the item because playing varying environment variation
Part goes tracking target to guarantee the real-time and robustness that track.
The present invention uses more algorithm fusion Moving Target Tracking Algorithms based on image classification, and reply have been directed to not well
Tracking problem caused by changing with factor, experiments verify that, the present invention is compared with other epidemic algorithms instantly in real-time and robust
It is all relatively high in property.
Sparse matrix is mathematically is defined as: if in a matrix, most elements is 0, this matrix is referred to as sparse square
Battle array (sparse matrix).Sparse representation method provides new thinking for image processing in recent years.Research shows that visual cortex
Using sparse coding principle, the sparse representation method based on sparse coding can preferably portray people for the expression of complex stimulus
Cognitive features of the class vision system to image.Sparse coding is carried out to a large amount of image feature information using sparse matrix, simultaneously
Dimensionality reduction compression is carried out to the vector information of sparse coding, algorithm is greatly reduced and obtains calculation amount, improve the operational efficiency of algorithm.
Mean shift algorithm is a kind of nonparametric technique based on density gradient, finds target position by interative computation,
Realize target following.The advantages of algorithm is that principle is simple, operational efficiency is high therefore wide to certain adaptive ability is blocked
General applies in target tracking domain.Meanwhile mean shift algorithm due to the calculating of itself be by calculate probability distribution density
Realize target following, therefore, when the size for tracking target is bigger, the more algorithms of the pixel quantity that count and calculate
Calculation amount is bigger, and efficiency is lower.Simultaneously as mean value drift algorithm is using the mode based on color characteristic, it is quick to illumination
Sense will lead to tracking drift in background situation similar with color of object.
There is a disadvantage according to mean shift algorithm, and combines the pretreatment classification of the image sequence of early period of the invention, it can be with
The tracking disadvantage for cleverly avoiding the problems such as mean shift algorithm is to illumination and similar background takes full advantage of it to blocking
Robustness and efficient operation efficiency.
Summary of the invention
A kind of method for tracking target for being designed to provide more algorithms cooperation based on image classification of the application, including
Step: (1) all target components are initialized;(2) judge whether current goal is front cross frame: if so, calculating using BW-SOAMS
Method tracks target;If it is not, inputting t (t > 2) frame image, then classify to image sequence judgement;(3) it is adjusted according to classification
With corresponding Moving Target Tracking Algorithm;(4) real-time update of target's feature-extraction, sample and object candidate area;(5) it gives
The position for tracking target out, reads in next frame, continues through b judgement classification tracking until the last frame of video sequence.
Preferably, the method that judgement is classified is carried out to image sequence in the step 2 are as follows:
P=N2/N1
Wherein, the total pixel number of image is defined as N, and the pixel of customized threshold binarization treatment changes number after frames differencing
For N1, the pixel variation number in target three times regional scope is N2, solves different zones pixel variation number respectively and accounts for global pixel
Specific gravity P.
Preferably, step 2 specific classification standard are as follows: it is locally interested that target is divided into image sequence by specific gravity P
Regional change and image overall variation, wherein the scale for incorporating and improving background weighting is called in the variation of target local region of interest
The Moving Target Tracking Algorithm of direction-adaptive;The Moving Target Tracking Algorithm based on compression tracking is called in image overall variation.
Preferably, the influence factor of the image of described image global change is illumination variation, context similarity and background
It is fuzzy.
Preferably, the influence factor of target local region of interest variation is the change in size, rotationally-varying of target
And block variation.
It is global to being divided into according to the early period of image sequence and processing classification according to the advantage and disadvantage present invention of sparse matrix
The video clip of image sequence carries out target following, and the advantages of sparse matrix indicates algorithm is adequately utilized, reduces target
The dimension calculated is tracked, real-time performance of tracking caused by the calculation amount of algorithm is excessive caused by greatly reducing because of image overall variation
Ungratified problem.Simultaneously because rarefaction representation algorithm exist when blocking occurs in target following or the adjacent frame position of target it is mobile compared with
It will lead to tracking drift when big, and draw this kind of variation by the image classification process of early period in algorithm of the invention
The shortcomings that being divided into the localized variation of image, dexterously having avoided the algorithm, the computational efficiency for having given full play to its algorithm are high
Advantage.Make algorithm that there is higher operational efficiency and higher robustness in object tracking process.
It should be appreciated that aforementioned description substantially and subsequent detailed description are exemplary illustration and explanation, it should not
As the limitation to the claimed content of the present invention.
Detailed description of the invention
With reference to the attached drawing of accompanying, the more purposes of the present invention, function and advantage are by the as follows of embodiment through the invention
Description is illustrated, in which:
Fig. 1 shows the flow chart of the method for tracking target of more algorithms cooperation according to the present invention based on image classification;
The method for tracking target for more algorithms cooperation that Fig. 2 shows according to the present invention based on image classification based on image
The sequence of video images figure of global change;
Fig. 3 show it is according to the present invention based on image classification more algorithms cooperation target tracking algorism based on image
The sequence of video images figure of localized variation;
Fig. 4 shows three times of target of the target tracking algorism of more algorithms cooperation according to the present invention based on image classification
Area image changes pixel number and image overall pixel number percentage change figure.
Specific embodiment
By reference to exemplary embodiment, the purpose of the present invention and function and the side for realizing these purposes and function
Method will be illustrated.However, the present invention is not limited to exemplary embodiment as disclosed below;Can by different form come
It is realized.The essence of specification is only to aid in those skilled in the relevant arts' Integrated Understanding detail of the invention.
Hereinafter, the embodiment of the present invention will be described with reference to the drawings.In the accompanying drawings, identical appended drawing reference represents identical
Or similar component or same or like step.
As shown in Figure 1, the target of more algorithms cooperation of the invention based on image classificationTrackingStep are as follows:
Step 101: initializing all target components;
Step 102: judging whether current goal is front cross frame: if so, entering step 106;If it is not, i.e. input t (t >
2) frame image then classifies to image sequence judgement;
According to one embodiment of present invention, the method that judgement is classified is carried out to image sequence in the step 102
Are as follows:
P=N2/N1
Wherein, the total pixel number of image is defined as N, and the pixel of customized threshold binarization treatment changes number after frames differencing
For N1, the pixel variation number in target three times regional scope is N2, solves different zones pixel variation number respectively and accounts for global pixel
Specific gravity P.
According to one embodiment of present invention, step 2 specific classification standard are as follows: by specific gravity P to image sequence point
For the variation of target local region of interest and image overall variation, wherein the variation of target local region of interest, which calls to incorporate, to be changed
The adaptive Moving Target Tracking Algorithm of the dimension weighted into background;The fortune based on compression tracking is called in image overall variation
Tracking of maneuvering target algorithm.
According to one embodiment of present invention, the influence factor of the image of described image global change is illumination variation, back
Scape similarity and blurred background.
According to one embodiment of present invention, the influence factor of the target local region of interest variation is the ruler of target
It is very little variation, rotationally-varying and block variation.
Step 103: corresponding Moving Target Tracking Algorithm is called according to classification;
Step 104: the real-time update of target's feature-extraction, sample and object candidate area;
Step 105: provide tracking target position, read in next frame, continue through step 102 judgement classification tracking until
The last frame of video sequence.
Step 106: using BW-SOAMS algorithm keeps track target.
Fig. 2 and Fig. 3 is respectively the base of the target tracking algorism of more algorithms cooperation according to the present invention based on image classification
In the sequence of video images figure of image overall variation and localized variation;
As shown, step are as follows:
A. pretreatment classification is carried out to image sequence;
The reason of according to causing pixel to change, carries out statistics calculating to the amount of pixels of variation, solves variation pixel and accounts for the overall situation
The specific gravity of pixel;
B. by making needle difference method between adjacent image frame, the image after obtaining frame difference;
C. the binary image for carrying out customized threshold value to image-region is handled;
Wherein, image sequence is divided by two classes according to the difference of the order of magnitude:
One kind is the variation based on image overall;Variation is mainly by illumination variation, context similarity and blurred background
It is caused;
Another kind of is the variation based on image local area-of-interest, and variation is mainly by the change in size of target, rotation
Change and blocks caused by variation.
Target is tracked according to processing Selection and call corresponding target tracking algorism.
The pixel variation for counting target three times region and image overall, is become using part and the global pixel number combined
Change to count can be effectively avoided and the phenomenon that being classified as global change of mistake is acutely led to by local pixel variation.
Image sequence pretreatment classification is the image by the frame difference operation between consecutive frame image, after obtaining frame difference.So
The image binaryzation for carrying out customized threshold value (threshold value is 0.2 in inventive algorithm) to image-region afterwards is handled.By testing
Card classifies to image sequence using the specific gravity that variation pixel accounts for global pixel, it is ensured that certain classification accuracy, but
It is, when a large amount of variations of target local region of interest pixel (are mainly seriously blocked by target itself or the size of target
Caused by serious variation occurs) when, it is above-mentioned simple using image will occur by the way of the specific gravity for changing pixel and accounting for global pixel
Classification error, mistake are divided into global change.Therefore, the present invention counts the pixel of target three times region and image overall
Variation, using part and the global pixel number variation statistics combined it is possible to prevente effectively from acutely leading to mistake by local pixel variation
The phenomenon that being mistakenly classified as global change.
Assuming that the total pixel number of image we be defined as N, image passes through the change of customized threshold binarization treatment after frame difference
Change pixel number is N1, and pixel variation number is N2 in the range of the three times area of target area, solves variation pixel respectively and accounts for global picture
The specific gravity of element.The present invention carries out verification experimental verification, figure to the representational image sequence segment of 20 the international standards organization standard generalized markup databases
Photo section and statistical conditions such as the following table 1.
The present invention changes number N1 in image overall pixel by comparing the pixel variation number N2 of target three times surface area, fixed
Adopted P is N2/N1, and by being compared statistics to above-mentioned 20 sections of sequence of pictures, the value range of P is 1 to 100%, and the present invention is with regard to P
Relationship of the value between the accuracy rate of image classification make corresponding curve, observation is schemed when P value is from 1% to 100%
As classification it is accurate between relation curve, the curve between accuracy rate and ratio is as shown in Figure 4.
Pass through the relation curve between accuracy rate and ratio P, it can be seen that the standard of image classification when P takes 10% or 11%
Really 100%, therefore we choose 10% as categorised demarcation line.The present invention is determined as using the local method in the overall situation is planned as a whole
Image local variation and global change.The case where effectively avoiding because of localized target region due to target itself (as: large area
Or all block, be change in size, rotationally-varying) acute variation and a possibility that being classified as global change of mistake, Ke Yizhun
Really divide the image into two classes.For the moving target for not selecting suitable robustness high with us for causing pixel variation principle
Track algorithm tracks moving target, to not only improve the operation efficiency of algorithm but also improve the robustness of algorithm.
Sequence of video images is accurately divided into two classes, the i.e. view based on image overall variation by the classification of draw above shape
Frequency image sequence and the sequence of video images changed based on image local are corresponding suitable according to different video sequence selections
Moving Target Tracking Algorithm.
The present invention uses the adaptive mean shift algorithm of dimension (referred to as: the BW-SOAMS) conduct weighted based on background
The correspondence track algorithm of local region of interest image sequence.Background weighting algorithm is fused in SOAMS algorithm, is effectively kept away
Exempted from by local region of interest context similarity variation, blurred background variation and background environment illumination it is slowly varying
It is classified as tracking drifting problem caused by the variation interested of part when identification and classification.While ensure that algorithm real-time, and protect
The high robust of target following is demonstrate,proved.
The invention discloses a kind of mobile target in complex background track algorithms.Due to target itself and environmental change
Diversity, leading to tracking, there are all difficulties in the process.Different Moving Target Tracking Algorithms respectively has excellent for different problems
Disadvantage, there are no a kind of target tracking algorisms that pervasive robustness is good up to now.Become for object variations and background environment
Change, proposes a kind of target tracking algorism of more calculations cooperation based on image classification.The variation of target itself causes part to feel emerging
The problem of interesting regional change be mainly by target object size, rotate and block caused by.It is complete caused by background environment variation
Office's variation issue is mainly as caused by illumination, context similarity and blurred background.The algorithm is first to image sequence pretreatment point
Class, then selection is suitble to track drifting problem caused by the environmental change.Experiments verify that the present invention is compared with other popular mesh
Marking track algorithm has better robustness.
Compared with the prior art, the target tracking algorism of more algorithms cooperation proposed by the present invention based on image classification has
Beneficial effect is embodied in:
(1) present invention is handled sequence of video images from the source of tracking image sequence, with according to because playing image
The difference of reason of changes in advance classifies to the target and background to be tracked before target following.
(2) target itself is tracked because the reason of playing environmental change can be mainly divided into two classes in object tracking process
The variation of variation and tracking environmental, in actual test we have found that because the factor for playing target Self-variation mainly has tracking mesh
Target blocks variation and change in size, because playing mainly being changed by illumination variation and context similarity for environmental change etc..According to two
Image is not divided into two classes: image overall variation and local sense with us because image change range by kind variation before tracking
Interest regional change.Person's appropriate algorithm is not then selected to carry out tracking and positioning to video sequence, in this way for because playing varying environment variation
Condition go tracking target guarantee tracking real-time and robustness.
In conjunction with the explanation and practice of the invention disclosed here, the other embodiment of the present invention is for those skilled in the art
It all will be readily apparent and understand.Illustrate and embodiment is regarded only as being exemplary, true scope of the invention and purport are equal
It is defined in the claims.
Claims (3)
1. a kind of method for tracking target of more algorithms cooperation based on image classification, comprising steps of
(1) all target components are initialized;
(2) judge whether current goal is front cross frame: if so, using BW-SOAMS algorithm keeps track target;If it is not, i.e. input the
Image after two frames then classifies to image sequence judgement;
(3) corresponding Moving Target Tracking Algorithm is called according to classification;
(4) real-time update of target's feature-extraction, sample and object candidate area;
(5) provide tracking target position, read in next frame, continue through judgement classification tracking until video sequence last
Frame;
The method classified is judged to image sequence in the step 2 are as follows:
P=N2/N1
Wherein, the total pixel number of image is defined as N, and the pixel variation number of customized threshold binarization treatment is N after frames differencing1,
Pixel variation number in target three times regional scope is N2, the specific gravity that different zones pixel variation number accounts for global pixel is solved respectively
P;
The variation of target local region of interest is divided into image sequence by specific gravity P and image overall changes, wherein target part
The Moving Target Tracking Algorithm that area-of-interest variation calls the dimension for incorporating improvement background weighting adaptive;Image overall
The Moving Target Tracking Algorithm based on compression tracking is called in variation.
2. according to the method described in claim 1, it is characterized by: the influence factor of the image of described image global change is light
According to variation, context similarity and blurred background.
3. according to the method described in claim 1, it is characterized by: the influence factor of target local region of interest variation
For the change in size, rotationally-varying and block variation of target.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610917111.8A CN106504269B (en) | 2016-10-20 | 2016-10-20 | A kind of method for tracking target of more algorithms cooperation based on image classification |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610917111.8A CN106504269B (en) | 2016-10-20 | 2016-10-20 | A kind of method for tracking target of more algorithms cooperation based on image classification |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106504269A CN106504269A (en) | 2017-03-15 |
CN106504269B true CN106504269B (en) | 2019-02-19 |
Family
ID=58318139
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610917111.8A Active CN106504269B (en) | 2016-10-20 | 2016-10-20 | A kind of method for tracking target of more algorithms cooperation based on image classification |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106504269B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101739551A (en) * | 2009-02-11 | 2010-06-16 | 北京智安邦科技有限公司 | Method and system for identifying moving objects |
CN101901354A (en) * | 2010-07-09 | 2010-12-01 | 浙江大学 | Method for detecting and tracking multi targets at real time in monitoring videotape based on characteristic point classification |
CN102236901A (en) * | 2011-06-30 | 2011-11-09 | 南京大学 | Method for tracking target based on graph theory cluster and color invariant space |
CN104992453A (en) * | 2015-07-14 | 2015-10-21 | 国家电网公司 | Target tracking method under complicated background based on extreme learning machine |
CN105654505A (en) * | 2015-12-18 | 2016-06-08 | 中山大学 | Collaborative tracking algorithm based on super-pixel and system thereof |
CN105913455A (en) * | 2016-04-11 | 2016-08-31 | 南京理工大学 | Local image enhancement-based object tracking method |
-
2016
- 2016-10-20 CN CN201610917111.8A patent/CN106504269B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101739551A (en) * | 2009-02-11 | 2010-06-16 | 北京智安邦科技有限公司 | Method and system for identifying moving objects |
CN101901354A (en) * | 2010-07-09 | 2010-12-01 | 浙江大学 | Method for detecting and tracking multi targets at real time in monitoring videotape based on characteristic point classification |
CN102236901A (en) * | 2011-06-30 | 2011-11-09 | 南京大学 | Method for tracking target based on graph theory cluster and color invariant space |
CN104992453A (en) * | 2015-07-14 | 2015-10-21 | 国家电网公司 | Target tracking method under complicated background based on extreme learning machine |
CN105654505A (en) * | 2015-12-18 | 2016-06-08 | 中山大学 | Collaborative tracking algorithm based on super-pixel and system thereof |
CN105913455A (en) * | 2016-04-11 | 2016-08-31 | 南京理工大学 | Local image enhancement-based object tracking method |
Also Published As
Publication number | Publication date |
---|---|
CN106504269A (en) | 2017-03-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Rao et al. | Selfie video based continuous Indian sign language recognition system | |
Yuan et al. | Deep smoke segmentation | |
Qiu et al. | Learning spatio-temporal representation with pseudo-3d residual networks | |
Stallkamp et al. | Man vs. computer: Benchmarking machine learning algorithms for traffic sign recognition | |
CN107330364B (en) | A kind of people counting method and system based on cGAN network | |
Chen et al. | A rain pixel recovery algorithm for videos with highly dynamic scenes | |
Dawson-Howe | A practical introduction to computer vision with opencv | |
Torralba et al. | Using the forest to see the trees: exploiting context for visual object detection and localization | |
Khare et al. | A novel character segmentation-reconstruction approach for license plate recognition | |
CN101520894B (en) | Method for extracting significant object based on region significance | |
CN110503613B (en) | Single image-oriented rain removing method based on cascade cavity convolution neural network | |
CN110263712B (en) | Coarse and fine pedestrian detection method based on region candidates | |
Žemgulys et al. | Recognition of basketball referee signals from real-time videos | |
CN107644426A (en) | Image, semantic dividing method based on pyramid pond encoding and decoding structure | |
CN101655983B (en) | Device and method for exacting dominant color | |
CN107967695A (en) | A kind of moving target detecting method based on depth light stream and morphological method | |
Tiwari et al. | Dynamic texture recognition using multiresolution edge-weighted local structure pattern | |
CN107633226A (en) | A kind of human action Tracking Recognition method and system | |
CN105512618B (en) | Video tracing method | |
Cygert et al. | Toward robust pedestrian detection with data augmentation | |
CN104376334B (en) | A kind of pedestrian comparison method of multi-scale feature fusion | |
Wang et al. | Background extraction based on joint gaussian conditional random fields | |
Ramirez-Alonso et al. | Temporal weighted learning model for background estimation with an automatic re-initialization stage and adaptive parameters update | |
CN107944437A (en) | A kind of Face detection method based on neutral net and integral image | |
Abdo et al. | Arabic alphabet and numbers sign language recognition |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |