CN108460786A - A kind of high speed tracking of unmanned plane spot - Google Patents
A kind of high speed tracking of unmanned plane spot Download PDFInfo
- Publication number
- CN108460786A CN108460786A CN201810092025.7A CN201810092025A CN108460786A CN 108460786 A CN108460786 A CN 108460786A CN 201810092025 A CN201810092025 A CN 201810092025A CN 108460786 A CN108460786 A CN 108460786A
- Authority
- CN
- China
- Prior art keywords
- target
- speed
- module
- tracking
- area
- 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
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- 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 relates to a kind of high speed trackings of unmanned plane spot, belong to image procossing, computer vision field.After operating personnel or other algorithms obtain rough target location, first, by target exact envelope extraction module, precision target frame is obtained；Secondly, by running high-speed target tracking module, position of the target in subsequent video sequence is obtained；Finally, target scale change information is obtained by dimensional variation detection module, and then corrects high-speed target tracking module, realize robust tracking of the track algorithm to target.The method of the invention solves the problems, such as the existing of UAV Video processing system target following：1）It is accurately tracked when target gives indefinite；2）The smaller texture unobvious of prime number shared by target picture to be tracked；3）Tracking robustness is insufficient when target scale changes；4）Real-time performance of tracking is poor under existing hardware platform.
Description
Technical field
The present invention relates to a kind of high speed trackings of unmanned plane spot, belong to image procossing, computer vision neck
Domain.
Background technology
Target following technology is widely used in unmanned plane operational detection, precision strike field, is also positioned for target, and detection is known
Not etc. information-handling techniques do not provide information support.The target tracking algorism of real-time robust, it is possible to reduce the behaviour of ground monitoring personnel
It bears, while being indispensable one of the step of unmanned plane automation and intelligentification.
At present in engineering application, correlation tracking algorithm is most widely used, due to being multiplied instead of spatial domain by frequency domain
Convolution, significantly improves the algorithm speed of service, while algorithm is realized simple, facilitates hardware realization, hardware process speed compared with
Soon, but the poor easy generation tracing positional of algorithm robustness in target deformation and dimensional variation drifts about.In addition some other
More common target tracking algorism has TLD (Tracking-Learning-Detection) algorithm, CT (Compressive
Tracking) algorithm, CSK (Exploiting the Circulant Structure of Tracking-by-Detection
With Kernels) algorithm, DSST (Accurate Scale Estimation for Robust Visual Tracking)
Algorithm, KCF (High-Speed Tracking with Kernelized Correlation Filters) algorithm, ACT
(Adaptive Color Attributes for Real-Time Visual Tracking) algorithm, LCT (Long-term
Correlation Tracking) algorithm etc..
TLD algorithms are by way of on-line study, the strategy combination intermediate value optical flow method using PN study and online cascade point
Class device is, it can be achieved that long-time target following, but it is low to track efficiency.Processing speed of the image of 320*240 resolution sizes in the later stage
Degree is only 5fps.CT algorithms generate a sparse projection matrix at random, and in low-dimensional feature space, directly application is by sparse
Compressive features after matrix projection classify to target, and algorithm tracking effect in target area texture-rich is preferable, can
Reach the speed of service of 30 frame per second or so；The property of CSK algorithm application circular matrixes, and it is to be detected by Fourier transform pairs
High speed SVM detections and training are realized in region, can reach the tracking velocity of 300 frame per second or so, but when target scale changes
It is not sufficiently effective.DSST algorithms propose an independent target scale detection module, which is advantageously integrated, while the speed of service
It is high.KCF algorithms and ACT algorithms are the improvement of CSK algorithms, by introducing FHOG features and the original CSK of ColorName matrix optimizings
Algorithm, algorithm have a certain upgrade, and also can reach real-time tracking result.
Algorithm above achieves good tracking effect in daily monitoring field, but scouts processing system in unmanned plane
In, due to taking photo by plane preferable promotion and application can not be obtained with the limitation of various conditions in operational environment.
Target following technology mainly faces problems in UAV Video processing system：
1) target specific mode can only when often operating due to the limitation of crawl target application environment when unmanned plane is scouted
The Position Approximate of target is clicked, accurate target rectangle frame can not be given；
2) in unmanned plane video, pixel number shared by target is insufficient, and ratio is smaller in entire picture, textural characteristics
Unobvious；
3) in unmanned plane video, since shooting image is loaded and the influence of aspect, lead to target angle ruler
Degree variation is apparent.
4) it on the basis of existing hardware is handled, needs to ensure real-time performance of tracking.
Invention content
The present invention provides a kind of high-speed target tracking method of unmanned plane video, plans as a whole performance and the adaptation of algorithm
Property, merge both at home and abroad existing algorithm the advantages of；Solve the problems, such as the existing of UAV Video processing system target following：1) mesh
Mark accurately tracks when given indefinite；2) the smaller texture unobvious of prime number shared by target picture to be tracked；3) when target scale changes
It is insufficient to track robustness；4) real-time performance of tracking is poor under existing hardware platform.
The present invention is achieved by the following technical solutions：
A kind of high speed tracking of unmanned plane spot, including：
After obtaining target rough position region, target exact envelope area or mesh are obtained by target exact envelope extraction module
Mark notable feature area；
Become according to target exact envelope area or target notable feature area initialization high-speed target tracking module and scale
Change detection module, high-speed target tracking module described in continuous service obtains position of the target in subsequent video sequence；
It obtains per behind target location in frame image, target in current frame image is obtained by the dimensional variation detection module
Exact scale corrects the high-speed target tracking module based on the target exact scale, to realize the robust tracking to target,
The problem of reducing tracker degeneration drift caused by changing due to target size.
Further, the method is specially：
Step1:Target rough position region is obtained by manual operation or other conventional algorithms；
Step2:Operational objective exact envelope extraction module obtains target exact envelope area or target notable feature area；
Step3:According to target exact envelope area or target notable feature area initialization high-speed target tracking module and
Dimensional variation detection module；
Step4:The high-speed target tracking module is run, the region of target in the current frame is obtained；
Step5:The dimensional variation detection module is run, target exact scale is obtained；
Step6:Detection zone is scaled according to the target exact scale, updates the high-speed target tracking module,
And the parameter of the update dimensional variation detection module；
Step7:Step4-Step6 is repeated, until video terminates.
Further, the target exact envelope extraction module carries out conspicuousness area near the rough target location
Domain is detected, and obtains target exact envelope area or target notable feature area, and the high-speed target tracking module is accurate with the target
Envelope area or target notable feature area represent target exact position into line trace, reduce background to greatest extent to the high speed mesh
The influence that mark tracking module is brought, and can be used in characterizing target area, the target exact envelope extraction module uses frequency spectrum
Residual error method SR carries out salient region detection.
Further, the high-speed target tracking module uses improved CSK algorithms into line trace, the improved CSK
Track algorithm is merged by extracting most notable color characteristic and gray feature, can on the basis of ensureing real-time, increase with
Track device working robust.
Further, the improved CSK track algorithms include the improvement to grader feature and change to more new strategy
Into；
Improved method to grader is：Notable Color Channel feature is introduced on the basis of original gray feature, is used
Feature used in the high-speed target tracking module described in hands-on；
Improved method to more new strategy is：Only meet centainly when the high-speed target tracking module obtains testing result
The high-speed target tracking module is updated when condition, is not otherwise updated.
Further, the acquisition of the notable Color Channel feature is specially：It is corresponding aobvious that area to be tested is obtained first
Color Channel is write, and the notable Color Channel feature is obtained to the notable Color Channel normalized；
Wherein, the acquisition methods of the notable Color Channel are：
(1) before high-speed target tracking module initialization, target area image RGB triple channel is detached, is obtained
The individual matrix in each channel；
(2) the matrix mean value for seeking each channel respectively judges that the maximum value in three mean values is denoted as the notable color
Channel.
Further, the size measurement module uses the size measurement module in DSST algorithms, and the size measurement
Module is designed based on online SVM classifier, is inputted and is instructed as training sample using each scale image around current goal when initial
Practice grader, in actual use after tracker obtains target location, current goal week is detected centered on current goal position
The confidence level for enclosing each scale image obtains accurate target scale information, last according to the target scale information detected
The high-speed target tracking module is corrected, to ensure the accuracy of the high-speed target tracking module.
Further, the method that the high-speed target tracking module is corrected by the dimensional variation detection module is specific
For：
The dimensional variation detection module has the scale of N number of candidate when detecting every time, for each candidate scale s ∈ S,
S is zoom factor, and zoom factor S is shown below：
Wherein a indicates that specific scaling cadence, n indicate that scaling multiplying power, N indicate candidate scale number；N is bigger, calculates the time
It is longer, but size measurement effect is better, and the smaller size measurement effects of N are poorer, but it is shorter to calculate the time.
An image block J is obtained centered on estimating central points, by all image block JsZoom to P × Q sizes, P and Q divide
Not Wei image width and height, extract corresponding HOG features, using the SVM models in DSST algorithms, calculate each image block
ResponseAnd then obtain optimum size size：
Wherein,Indicate the optimum size size obtained,Expression takes the maximum value pair of each image block response
The scale answered is as optimum size size；
After determining optimum size size, according to its corresponding zoom factor s, original image is zoomed in and out, and with contracting
Area update high speed tracker after putting, reaches to the modified purpose of high speed tracker.
The advantageous effects of the present invention：
1) the method for the invention operational efficiency is high, and under i7-4790 processors, 8GB memory conditions, single goal is averagely transported
Scanning frequency degree can carry out multiple target and handle in real time up to 200 frames/second.
2) the method for the invention is accurately wrapped by obtaining target using conspicuousness detection algorithm in tracking pretreatment stage
Network or target notable feature region into line trace, reduce due to target given accuracy it is insufficient in the case of cause tracker to be degenerated drift
The problem of shifting.
3) the method for the invention improves traditional C/S K algorithms, on the basis of ensureing algorithm performs real-time, carries as possible
Rise algorithm running precision.
4) the method for the invention by merge change of scale detector, can precise real-time detection target size, reduce by
Caused by target size changes the problem of tracker degeneration drift.
Description of the drawings
Fig. 1 is the overall operation block diagram of the high speed tracking of unmanned plane spot in the embodiment of the present invention.
Fig. 2 is target exact envelope extraction effect figure in the embodiment of the present invention.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is explained in further detail.It should be appreciated that specific embodiment described herein is used only for explaining the present invention, and
It is not used in the restriction present invention.
On the contrary, the present invention covers any replacement done in the spirit and scope of the present invention being defined by the claims, repaiies
Change, equivalent method and scheme.Further, in order to make the public have a better understanding the present invention, below to the thin of the present invention
It is detailed to describe some specific detail sections in section description.Part without these details for a person skilled in the art
Description can also understand the present invention completely.
Embodiment 1
The present embodiment provides a kind of high speed tracking of unmanned plane spot, the method flow diagram such as Fig. 1 institutes
Show, the method on the basis of artificial given or other algorithms calculate and obtain target rough position region, passes through target first
Exact envelope extraction module obtains target exact envelope area or target notable feature area；Secondly with the target exact envelope area or
Target notable feature area initializes high-speed target tracking module and dimensional variation detection module, high-speed target described in continuous service with
Track module obtains position of the target in subsequent video sequence；It obtains per behind target location in frame image, is become by the scale
Change detection module and obtains target exact scale in current frame image, and the testing result amendment update after being determined according to exact scale
Tracker.
The method is specially：
Step1:Target rough position region is obtained by manual operation or other conventional algorithms；
Step2:Operational objective exact envelope extraction module obtains target exact envelope area or target notable feature area；
Step3:According to target exact envelope area or target notable feature area initialization high-speed target tracking module and
Dimensional variation detection module；
Step4:The high-speed target tracking module is run, the region of target in the current frame is obtained；
Step5:The dimensional variation detection module is run, target exact scale is obtained；
Step6:Detection zone is scaled according to the target exact scale, updates the high-speed target tracking module,
And the parameter of the update dimensional variation detection module；
Step7:Step4-Step6 is repeated, until video terminates.
The target exact envelope extraction module in the present embodiment pays close attention to the notable of mechanism using based on human eye vision
Property detection algorithm, carry out salient region using spectral residuum method (Spectral Residual, SR) in the present embodiment and carry
It takes, but is not limited to spectral residuum algorithm, may be used in actual use according to unmanned plane carry video camera shooting effect difference
Other salient region detection algorithms.The target exact envelope extraction module is shown near the rough target location
Work property region detection obtains target exact envelope area or target notable feature area, and aobvious with the target exact envelope area or target
It writes characteristic area and represents target exact position into line trace, reduce the influence that background brings tracker to greatest extent, and can
For characterizing target area.
The high-speed target tracking module using improved CSK algorithms into line trace, the improved CSK track algorithms,
By extracting most notable color characteristic and gray feature fusion, tracker work can be increased on the basis of ensureing real-time
Robustness.
In the present embodiment, improved CSK track algorithms include the improvement to grader feature and the improvement to more new strategy
Two parts：
1) it is to the improved method of grader：Notable Color Channel feature is introduced on the basis of original gray feature, is obtained
For feature used in high-speed target tracking module described in hands-on；Specially：
According to the coloured image of area to be tested, gray level image grayImg is generated, place is normalized to the gray level image
Reason obtains gray feature grayMat=grayImg/255-0.5；Obtain the corresponding notable color channel image of area to be tested
SaColorImg, and notable Color Channel feature saColorMat=is obtained to the notable color channel image normalized
saColorImg/255–0.5；Since the value range of grayImg and saColorImg elements is 0~255.GrayMat at this time
Value range with saColorMat is -0.5~0.5；For special used in high-speed target tracking module described in hands-on
FeatMat=[grayMat, saColorMat] is levied, wherein [] indicates that gray feature and notable Color Channel feature are respectively real
Border uses two channels of eigenmatrix.
Wherein, the acquisition methods of the notable color channel image are：
Before high-speed target tracking module initialization, target area image RGB triple channel is detached, is obtained every
A individual matrix in channel.I.e.：ImgColor={ ImgR, ImgG, ImgB }, wherein ImgColor indicate target area cromogram
Picture, ImgR indicate that target area red channel matrix, ImgG indicate that target area green channel matrix, ImgB indicate target area
Blue channel matrix.Each access matrix mean value is sought respectively, the mean value of redgreenblue access matrix is denoted as meanR respectively,
meanG,meanB.Judge that the maximum value in three mean values is denoted as notable Color Color channel image, and records notable Color Color
Channel image is saColorImg.
2) it is to the improved method of more new strategy：The reliability for judging the high-speed target tracking module tracking result,
The high-speed target tracking module model is updated when meeting following two condition, is not otherwise updated；
Condition 1：Fmax>Fth
Condition 2：Fmax/Fmean>Fsth
Wherein Fmax indicates that the maximum response in high-speed target tracking module current detection result, Fmean indicate current
The corresponding mean value of high-speed target tracking module result, Fth expressions pre-set high-speed target tracking module response threshold value, and Fsth is pre-
First setting indicates high-speed target tracking module testing result conspicuousness threshold value, and Fth takes 0.65, Fsth to take 115 in the present embodiment.
The size measurement module uses the size measurement module in DSST algorithms, and the size measurement module is based on
Line SVM classifier designs, and training grader is inputted as training sample using each scale image around current goal when initial, real
When border uses after tracker obtains target location, each scale image around current goal is detected centered on current goal position
Confidence level, obtain accurate target scale information, it is last according to high speed described in the target scale Information revision detected
Target tracking module, to ensure the accuracy of tracking result.
The dimensional variation detection module has the scale of N number of candidate when detecting every time, for each candidate scale s ∈ S,
S is zoom factor, and zoom factor S is shown below：
Wherein a indicates that specific scaling cadence, n indicate that scaling multiplying power, N indicate candidate scale number.N is bigger, calculates the time
It is longer, but size measurement effect is better, and the smaller size measurement effects of N are poorer, but it is shorter to calculate the time.
In the present embodiment, a values 1.02, n values 33；
An image block J is obtained centered on estimating central points, by all image block JsZoom to P × Q sizes, P and Q divide
Not Wei image width and height, extract corresponding HOG features, be herein conventional DSST algorithms using the SVM models in DSST algorithms
In SVM models, do not improve, calculate the response of each image blockAnd then obtain optimum size size：
Wherein,Indicate the optimum size size obtained,Expression takes the maximum value pair of each image block response
The scale answered is as optimum size size；
After determining optimum size size, according to its corresponding zoom factor s, original image is zoomed in and out, and with contracting
Area update high speed tracker after putting, reaches to the modified purpose of high speed tracker.
The technology of the present invention solves the problems, such as：The characteristics of scouting video sequence based on unmanned plane and difficult point, plan as a whole the property of algorithm
Can and adaptability, a kind of the advantages of merging existing algorithm both at home and abroad, it is proposed that high speed tracking suitable for unmanned plane spot
Method solves the problems of UAV Video processing system target following, includes mainly：
1) the problem of being accurately tracked when target gives indefinite.
2) prime number shared by target picture to be tracked is smaller, the unconspicuous problem of texture.
3) problem of robustness deficiency is tracked when target scale changes.
4) under existing hardware platform the problem of real-time performance of tracking.
Claims (8)
1. a kind of high speed tracking of unmanned plane spot, which is characterized in that
After obtaining target rough position region, target exact envelope area is obtained by target exact envelope extraction module or target is aobvious
Write characteristic area；
According to target exact envelope area or target notable feature area initialization high-speed target tracking module and dimensional variation inspection
Module is surveyed, high-speed target tracking module described in continuous service obtains position of the target in subsequent video sequence；
It obtains per behind target location in frame image, it is accurate that target in current frame image is obtained by the dimensional variation detection module
Scale corrects the high-speed target tracking module based on the target exact scale, to realize the robust tracking to target, reduces
Caused by changing due to target size the problem of tracker degeneration drift.
2. a kind of high speed tracking of unmanned plane spot according to claim 1, which is characterized in that the method has
Body is：
Step1:Target rough position region is obtained by manual operation or other conventional algorithms；
Step2:Operational objective exact envelope extraction module obtains target exact envelope area or target notable feature area；
Step3:According to target exact envelope area or target notable feature area initialization high-speed target tracking module and scale
Change detection module；
Step4:The high-speed target tracking module is run, the region of target in the current frame is obtained；
Step5:The dimensional variation detection module is run, target exact scale is obtained；
Step6:Detection zone is scaled according to the target exact scale, updates the high-speed target tracking module, and
Update the parameter of the dimensional variation detection module；
Step7:Step4-Step6 is repeated, until video terminates.
3. a kind of high speed tracking of unmanned plane spot according to claim 1 or claim 2, which is characterized in that the mesh
Exact envelope extraction module is marked, near the rough target location, salient region detection is carried out, obtains target exact envelope
Area or target notable feature area, the high-speed target tracking module are represented with the target exact envelope area or target notable feature area
The influence that background brings the high-speed target tracking module, and energy are reduced to greatest extent into line trace in target exact position
It is enough in characterization target area, the target exact envelope extraction module carries out salient region inspection using spectral residuum method SR
It surveys.
4. a kind of high speed tracking of unmanned plane spot according to claim 1 or claim 2, which is characterized in that the height
For fast target tracking module using improved CSK algorithms into line trace, the improved CSK track algorithms are most notable by extracting
Color characteristic and gray feature fusion can increase tracker working robust on the basis of ensureing real-time.
5. a kind of high speed tracking of unmanned plane spot according to claim 4, which is characterized in that described improved
CSK track algorithms include the improvement to grader feature and the improvement to more new strategy；
Improved method to grader is：Notable Color Channel feature is introduced on the basis of original gray feature, is obtained for real
Train feature used in the high-speed target tracking module in border；
Improved method to more new strategy is：Only meet certain condition when the high-speed target tracking module obtains testing result
High-speed target tracking module described in Shi Gengxin, does not otherwise update.
6. a kind of high speed tracking of unmanned plane spot according to claim 5, which is characterized in that the notable face
The acquisition of chrominance channel feature is specially：The corresponding notable Color Channel of area to be tested is obtained first, and logical to the notable color
Road normalized obtains the notable Color Channel feature；
Wherein, the acquisition methods of the notable Color Channel are：
(1) before high-speed target tracking module initialization, target area image RGB triple channel is detached, is obtained each
The individual matrix in channel；
(2) the matrix mean value for seeking each channel respectively judges that the maximum value in three mean values is denoted as the notable Color Channel.
7. a kind of high speed tracking of unmanned plane spot according to claim 1 or claim 2, which is characterized in that the ruler
Detection module is spent using the size measurement module in DSST algorithms, and the size measurement module is set based on online SVM classifier
Meter, using each scale image around current goal as the trained grader of training sample input when initial, in actual use with
After track device obtains target location, the confidence level of each scale image around current goal is detected centered on current goal position, is obtained
Accurate target scale information is taken, it is last according to high-speed target tracking mould described in the target scale Information revision detected
Block, to ensure the accuracy of the high-speed target tracking module.
8. a kind of high speed tracking of unmanned plane spot according to claim 7, which is characterized in that pass through the ruler
Spending the method that change detection module corrects the high-speed target tracking module is specially：
The dimensional variation detection module has the scale of N number of candidate when detecting every time, be for each candidate scale s ∈ S, S
Zoom factor, zoom factor S are shown below：
Wherein, a indicates that specific scaling cadence, n indicate that scaling multiplying power, N indicate candidate scale number；
An image block J is obtained centered on estimating central points, by all image block JsZoom to P × Q sizes, P and Q are respectively
The width and height of image extract corresponding FHOG features, calculate the response of each image blockAnd then it is big to obtain optimum size
It is small：
Wherein,Indicate the optimum size size obtained,Expression takes the corresponding ruler of maximum value of each image block response
Degree is used as optimum size size；
After determining optimum size size, according to its corresponding zoom factor s, original image is zoomed in and out, and with scaling after
Area update high speed tracker, reach to the modified purpose of high speed tracker.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810092025.7A CN108460786A (en) | 2018-01-30 | 2018-01-30 | A kind of high speed tracking of unmanned plane spot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810092025.7A CN108460786A (en) | 2018-01-30 | 2018-01-30 | A kind of high speed tracking of unmanned plane spot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108460786A true CN108460786A (en) | 2018-08-28 |
Family
ID=63239352
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810092025.7A Pending CN108460786A (en) | 2018-01-30 | 2018-01-30 | A kind of high speed tracking of unmanned plane spot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108460786A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635649A (en) * | 2018-11-05 | 2019-04-16 | 航天时代飞鸿技术有限公司 | A kind of high speed detection method and system of unmanned plane spot |
| CN111291630A (en) * | 2020-01-17 | 2020-06-16 | 天津大学 | Long-term target tracking algorithm based on joint prediction-detection-correction framework |
Citations (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101996310A (en) * | 2009-08-12 | 2011-03-30 | Tcl数码科技（深圳）有限责任公司 | Face detection and tracking method based on embedded system |
| CN102156991A (en) * | 2011-04-11 | 2011-08-17 | 上海交通大学 | Quaternion based object optical flow tracking method |
| CN102184551A (en) * | 2011-05-10 | 2011-09-14 | 东北大学 | Automatic target tracking method and system by combining multi-characteristic matching and particle filtering |
| CN102929288A (en) * | 2012-08-23 | 2013-02-13 | 山东电力集团公司电力科学研究院 | Unmanned aerial vehicle inspection head control method based on visual servo |
| CN103035013A (en) * | 2013-01-08 | 2013-04-10 | 东北师范大学 | Accurate moving shadow detection method based on multi-feature fusion |
| CN103065331A (en) * | 2013-01-15 | 2013-04-24 | 南京工程学院 | Target tracking method based on correlation of space-time-domain edge and color feature |
| CN103745203A (en) * | 2014-01-15 | 2014-04-23 | 南京理工大学 | Visual attention and mean shift-based target detection and tracking method |
| CN105844647A (en) * | 2016-04-06 | 2016-08-10 | 哈尔滨伟方智能科技开发有限责任公司 | Kernel-related target tracking method based on color attributes |
| CN106204638A (en) * | 2016-06-29 | 2016-12-07 | 西安电子科技大学 | A kind of based on dimension self-adaption with the method for tracking target of taking photo by plane blocking process |
| CN106296742A (en) * | 2016-08-19 | 2017-01-04 | 华侨大学 | A kind of online method for tracking target of combination Feature Points Matching |
| CN106599836A (en) * | 2016-12-13 | 2017-04-26 | 北京智慧眼科技股份有限公司 | Multi-face tracking method and tracking system |
| CN106600572A (en) * | 2016-12-12 | 2017-04-26 | 长春理工大学 | Adaptive low-illumination visible image and infrared image fusion method |
| CN106683110A (en) * | 2015-11-09 | 2017-05-17 | 展讯通信(天津)有限公司 | User terminal and object tracking method and device thereof |
| CN106709472A (en) * | 2017-01-17 | 2017-05-24 | 湖南优象科技有限公司 | Video target detecting and tracking method based on optical flow features |
| CN106846377A (en) * | 2017-01-09 | 2017-06-13 | 深圳市美好幸福生活安全***有限公司 | A kind of target tracking algorism extracted based on color attribute and active features |
| CN106874854A (en) * | 2017-01-19 | 2017-06-20 | 西安电子科技大学 | Unmanned plane wireless vehicle tracking based on embedded platform |
| CN106887011A (en) * | 2017-01-20 | 2017-06-23 | 北京理工大学 | A kind of multi-template method for tracking target based on CNN and CF |
| CN106952294A (en) * | 2017-02-15 | 2017-07-14 | 北京工业大学 | A kind of video tracing method based on RGB D data |
| CN106997597A (en) * | 2017-03-22 | 2017-08-01 | 南京大学 | It is a kind of based on have supervision conspicuousness detection method for tracking target |
| CN107169994A (en) * | 2017-05-15 | 2017-09-15 | 上海应用技术大学 | Correlation filtering tracking based on multi-feature fusion |
| CN107358623A (en) * | 2017-07-12 | 2017-11-17 | 武汉大学 | A kind of correlation filtering track algorithm based on conspicuousness detection and robustness size estimation |
| CN107423702A (en) * | 2017-07-20 | 2017-12-01 | 西安电子科技大学 | Video target tracking method based on TLD tracking systems |
| CN107481264A (en) * | 2017-08-11 | 2017-12-15 | 江南大学 | A kind of video target tracking method of adaptive scale |
| CN107564022A (en) * | 2017-07-13 | 2018-01-09 | 西安电子科技大学 | Saliency detection method based on Bayesian Fusion |
-
2018
- 2018-01-30 CN CN201810092025.7A patent/CN108460786A/en active Pending
Patent Citations (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101996310A (en) * | 2009-08-12 | 2011-03-30 | Tcl数码科技（深圳）有限责任公司 | Face detection and tracking method based on embedded system |
| CN102156991A (en) * | 2011-04-11 | 2011-08-17 | 上海交通大学 | Quaternion based object optical flow tracking method |
| CN102184551A (en) * | 2011-05-10 | 2011-09-14 | 东北大学 | Automatic target tracking method and system by combining multi-characteristic matching and particle filtering |
| CN102929288A (en) * | 2012-08-23 | 2013-02-13 | 山东电力集团公司电力科学研究院 | Unmanned aerial vehicle inspection head control method based on visual servo |
| CN103035013A (en) * | 2013-01-08 | 2013-04-10 | 东北师范大学 | Accurate moving shadow detection method based on multi-feature fusion |
| CN103065331A (en) * | 2013-01-15 | 2013-04-24 | 南京工程学院 | Target tracking method based on correlation of space-time-domain edge and color feature |
| CN103745203A (en) * | 2014-01-15 | 2014-04-23 | 南京理工大学 | Visual attention and mean shift-based target detection and tracking method |
| CN106683110A (en) * | 2015-11-09 | 2017-05-17 | 展讯通信(天津)有限公司 | User terminal and object tracking method and device thereof |
| CN105844647A (en) * | 2016-04-06 | 2016-08-10 | 哈尔滨伟方智能科技开发有限责任公司 | Kernel-related target tracking method based on color attributes |
| CN106204638A (en) * | 2016-06-29 | 2016-12-07 | 西安电子科技大学 | A kind of based on dimension self-adaption with the method for tracking target of taking photo by plane blocking process |
| CN106296742A (en) * | 2016-08-19 | 2017-01-04 | 华侨大学 | A kind of online method for tracking target of combination Feature Points Matching |
| CN106600572A (en) * | 2016-12-12 | 2017-04-26 | 长春理工大学 | Adaptive low-illumination visible image and infrared image fusion method |
| CN106599836A (en) * | 2016-12-13 | 2017-04-26 | 北京智慧眼科技股份有限公司 | Multi-face tracking method and tracking system |
| CN106846377A (en) * | 2017-01-09 | 2017-06-13 | 深圳市美好幸福生活安全***有限公司 | A kind of target tracking algorism extracted based on color attribute and active features |
| CN106709472A (en) * | 2017-01-17 | 2017-05-24 | 湖南优象科技有限公司 | Video target detecting and tracking method based on optical flow features |
| CN106874854A (en) * | 2017-01-19 | 2017-06-20 | 西安电子科技大学 | Unmanned plane wireless vehicle tracking based on embedded platform |
| CN106887011A (en) * | 2017-01-20 | 2017-06-23 | 北京理工大学 | A kind of multi-template method for tracking target based on CNN and CF |
| CN106952294A (en) * | 2017-02-15 | 2017-07-14 | 北京工业大学 | A kind of video tracing method based on RGB D data |
| CN106997597A (en) * | 2017-03-22 | 2017-08-01 | 南京大学 | It is a kind of based on have supervision conspicuousness detection method for tracking target |
| CN107169994A (en) * | 2017-05-15 | 2017-09-15 | 上海应用技术大学 | Correlation filtering tracking based on multi-feature fusion |
| CN107358623A (en) * | 2017-07-12 | 2017-11-17 | 武汉大学 | A kind of correlation filtering track algorithm based on conspicuousness detection and robustness size estimation |
| CN107564022A (en) * | 2017-07-13 | 2018-01-09 | 西安电子科技大学 | Saliency detection method based on Bayesian Fusion |
| CN107423702A (en) * | 2017-07-20 | 2017-12-01 | 西安电子科技大学 | Video target tracking method based on TLD tracking systems |
| CN107481264A (en) * | 2017-08-11 | 2017-12-15 | 江南大学 | A kind of video target tracking method of adaptive scale |
Non-Patent Citations (1)
| Title |
|---|
| MARTIN DANELLJAN等: "Adaptive Color Attributes for Real-Time Visual Tracking", 《CVPR》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635649A (en) * | 2018-11-05 | 2019-04-16 | 航天时代飞鸿技术有限公司 | A kind of high speed detection method and system of unmanned plane spot |
| CN111291630A (en) * | 2020-01-17 | 2020-06-16 | 天津大学 | Long-term target tracking algorithm based on joint prediction-detection-correction framework |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Boltes et al. | Automatic extraction of pedestrian trajectories from video recordings | |
| CN110276264B (en) | Crowd density estimation method based on foreground segmentation graph | |
| CN113674416B (en) | Three-dimensional map construction method and device, electronic equipment and storage medium | |
| CN110097536A (en) | Hexagon bolt looseness detection method based on deep learning and Hough transformation | |
| CN104392416B (en) | Video stitching method for sports scene | |
| CN104700404A (en) | Fruit location identification method | |
| CN108197604A (en) | Fast face positioning and tracing method based on embedded device | |
| CN113435282B (en) | Unmanned aerial vehicle image ear recognition method based on deep learning | |
| Zhang et al. | Application of migration image registration algorithm based on improved SURF in remote sensing image mosaic | |
| CN106709938B (en) | Based on the multi-target tracking method for improving TLD | |
| CN105279771B (en) | A kind of moving target detecting method based on the modeling of online dynamic background in video | |
| CN105469427B (en) | One kind is for method for tracking target in video | |
| CN109253722A (en) | Merge monocular range-measurement system, method, equipment and the storage medium of semantic segmentation | |
| CN107944437B (en) | A kind of Face detection method based on neural network and integral image | |
| CN108009556A (en) | A kind of floater in river detection method based on fixed point graphical analysis | |
| Ospina et al. | Simultaneous mapping and crop row detection by fusing data from wide angle and telephoto images | |
| CN105809673A (en) | SURF (Speeded-Up Robust Features) algorithm and maximal similarity region merging based video foreground segmentation method | |
| Chen et al. | A color-guided, region-adaptive and depth-selective unified framework for Kinect depth recovery | |
| CN108460786A (en) | A kind of high speed tracking of unmanned plane spot | |
| Low et al. | Robust e-nerf: Nerf from sparse & noisy events under non-uniform motion | |
| Shi et al. | A method for detecting pedestrian height and distance based on monocular vision technology | |
| CN115880683B (en) | Urban waterlogging ponding intelligent water level detection method based on deep learning | |
| CN116630352A (en) | Rock core measurement method and device based on bidirectional cascade pixel differential network | |
| CN103236053B (en) | A kind of MOF method of moving object detection under mobile platform | |
| CN110120030A (en) | Processing method, application, computer-readable medium and the disease occurring area measuring method of wheat diseases generation image |
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: 20180828 |