CN106780555A - A kind of sane motion target tracking method of high speed - Google Patents
A kind of sane motion target tracking method of high speed Download PDFInfo
- Publication number
- CN106780555A CN106780555A CN201611095524.9A CN201611095524A CN106780555A CN 106780555 A CN106780555 A CN 106780555A CN 201611095524 A CN201611095524 A CN 201611095524A CN 106780555 A CN106780555 A CN 106780555A
- Authority
- CN
- China
- Prior art keywords
- target
- tracking
- motion
- judge whether
- tracked
- 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 OR CALCULATING; 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
- Image Analysis (AREA)
Abstract
The invention provides a kind of sane motion target tracking method of high speed, it is characterised in that including step：A) motion detection is carried out to two field picture；B) the big target of motion and the Small object of motion are rejected；C) determine whether that two targets intersect, if not having, into step d, otherwise, moving Object Segmentation is opened according to the target for tracking, enter back into step d；D) Color Names features are extracted and metric transformation is carried out, optimal result is found；E) judge whether the target for having tracked has the target that can be matched；If no, into step f, otherwise into step g；F) judge whether to meet establishment fresh target condition；If not meeting, into step g, if meeting, fresh target is created, enter back into step g；G) target for tracking is updated；H) first visual angle is processed；I) judge whether rectangle frame extracts completely, if extracting completely, into next two field picture, otherwise, return to step c has continued to determine whether two target crossing instances.
Description
Technical field
The present invention relates to the method for motion target tracking in safety-protection system field of video monitoring.
Background technology
In recent years, the continuous attention with people to society to public safety, in video monitoring, either to target
Real-time tracking, or movement locus to target after it there is accident is analyzed, and or does other based on tracking
Intelligent algorithm, a sane Moving Target Tracking Algorithm of high speed is essential.
Kalman filter is the traditional method of a comparing, but it is linear, noise that the premise of Kalman filter is system
In Gaussian Profile, posterior probability is also Gaussian.Although the algorithm speed ratio is very fast, tracking effect is general, works as target
When number is relatively more, intersecting occurs in target, tracking effect is just especially unsatisfactory, and compares the result for relying on background modeling.
Recent years also emerges many track algorithms, and these algorithms also achieve some good effects, but still
Come with some shortcomings, for example, target occur dimensional variation, quickly move, go out visual angle, intersect when, tracking effect is just
Will not be highly desirable.
What the application was proposed is a kind of using colouring information Color Names features, then by metric transformation, is calculated
Euclidean distance, finds a target for matching the most, if due to motion detection effect or block etc. that factor causes it is short
Temporary target is lost, then carry out location estimating according to the last result for matching again, according to previous if running into target and intersecting
Individual motion result is split, if running into visual angle, the part Color Names features for extracting previous tracking result are entered
Row matching, finally realizes the sane tracking of high speed.
The content of the invention
The purpose of the application there are provided a kind of sane motion target tracking method of high speed, it is characterised in that including
Step：A) motion detection is carried out to two field picture；B) the big target of motion and the Small object of motion are rejected；C) two are determined whether
Target intersects, if not having, into step d, otherwise, moving Object Segmentation is opened according to the target for tracking, and enters back into step
Rapid d；D) according to the size for tracking target, Color Names are extracted in the nearest position of distance original target in intersection region
Feature simultaneously carries out metric transformation, calculates and track the Euclidean distance of target afterwards, finds optimal result；E) judge to have tracked
Target whether have the target that can be matched；If no, into step f, otherwise into step g；F) judge whether to meet wound
Build fresh target condition；If not meeting, into step g, if meeting, fresh target is created, enter back into step g；G) update with
The target of track；H) first visual angle is processed；I) judge whether rectangle frame extracts completely, if extracting completely, into next frame figure
Picture, otherwise, return to step c has continued to determine whether two target crossing instances.
Preferably, judge that the method whether two targets intersect is in the step c)：Two of previous frame image or
Multiple target ranges are close, and have close trend.
Preferably, just the processing method at visual angle is in the step h)：First visual angle concentrates on the marginal portion of image, first
Judge its edge which side is in, the relative position Color Names features for tracking target then found according to its position,
Then metric transformation is carried out, Euclidean distance is calculated, is then same target less than threshold value, continue to track, until this target mistake
Untill small disappearance.
Preferably, methods described sets a maximum frame number for allowing to lose, when utilization ColorNames features cannot
When matching the target for tracking, then temporarily its status indication to lose, when the target as matching is found again, then
According to the front and rear target location for detecting twice, position mean is calculated, added among the target for tracking；If even
The continuous frame number lost is more than or equal to the max-thresholds N for allowing to lose, then stop tracking target.
It should be appreciated that foregoing description substantially and follow-up description in detail are exemplary illustration and explanation, should not
As the limitation to claimed content of the invention.
Brief description of the drawings
With reference to the accompanying drawing enclosed, the present invention more purpose, function and advantages are by by the as follows of embodiment of the present invention
Description is illustrated, wherein：
Fig. 1 shows the flow chart of motion target tracking method of the invention；
Fig. 2 shows the treatment strategic process figure of the target transient loss of motion target tracking method of the invention.
Specific embodiment
By reference to one 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 one exemplary embodiment as disclosed below；Can by multi-form come
It is realized.The essence of specification is only to aid in various equivalent modifications Integrated Understanding detail of the invention.
Hereinafter, embodiments of the invention will be described with reference to the drawings.In the accompanying drawings, identical reference represents identical
Or similar part, or same or like step.
This patent proposes one kind and utilizes colouring information Color Names features, then by metric transformation, calculates Europe
Formula distance, finds a target for matching the most, if due to motion detection effect or block etc. that factor causes it is of short duration
Target lose, then location estimating is carried out according to the last result for matching again, according to previous if running into target and intersecting
Motion result is split, if running into visual angle, the part Color Names features for extracting previous tracking result are carried out
Matching, finally realizes the sane tracking of high speed.
Color Names are a kind of naming methods of color, and it is empty that the image of rgb space is mapped to Color Names by it
Between (Color Names spaces are 11 passages, are respectively black, blue, brown, grey, green, orange, pink,
Purple, red, white, yellow).It is excellent in images match that substantial amounts of test data indicates Color Names features
More property.
For the calculating of Euclidean distance,If characteristic dimension is excessive, for the meter of Euclidean distance
The complexity of calculation program can be greatly improved, and research shows, a suitable module significantly upper can improve algorithm
Performance.We can learn a module：
Wherein, dmnRepresentative feature vector xm, xnEuclidean distance, xmiRepresent characteristic vector xmThe dimension of i-th dimension, x, y point
Two characteristic vectors of Euclidean distance to be solved are not represented, and T represents the dimension of x, y, and A is coefficient matrix.
In order to solve an optimal module, it is contemplated that following optimization problem：
Wherein S is label identical sample binary groups, and D is the different sample binary groups of label, and this is a convex optimization
Problem, can be solved using the standard method of convex optimization problem.
The overall procedure of pursuit movement target proposed by the present invention is after having a new two field picture to enter algorithm, first
Carry out motion detection, for motion detection to target carry out a goal filtering first, reject excessive and too small target.
First once whether might have the judgement of intersection afterwards, intersect the basis for judging be two of former frame or
Person's multiple target range is close, and has close trend, if it is decided that be the target intersected, then basis tracks the big of target
It is small, extract Color Names features and carry out metric transformation in the nearest position of the former target of distance in intersection region, calculate afterwards
Euclidean distance with target is tracked, finds optimal result.
For the judgement of fresh target, be can not find in information in rectangle frame and the target for tracking matching any
One target can not horse back be considered as new target, but the continuous multiple frames after all detect it is matched
As a result.
Followed by going out the treatment at visual angle, the situation for going out visual angle typically all concentrates on the marginal portion of image, first determines whether
Its edge which side is in, then finds the relative position Color Names features for tracking target, then according to its position
Metric transformation is carried out, Euclidean distance is calculated, is then same target less than threshold value, continue to track, it is known that this target is too small to disappear
Untill mistake.
Specific steps as shown in figure 1, Fig. 1 shows the flow chart of motion target tracking method of the invention, including
Step：
Step 101：Motion detection is carried out to two field picture；
Step 102：Reject the big target of motion and the Small object of motion；
Step 103：Determine whether that two targets intersect, if not having, into step 105, otherwise, progressive step 104；
According to one embodiment of present invention, judge that the method whether two targets intersect is in the step 103：It is previous
Two or more target ranges of two field picture are close, and have close trend.
Step 104：Target according to tracking opens moving Object Segmentation, enters back into step 105；
Step 105：According to the size for tracking target, the nearest position of distance original target is extracted in intersection region
Color Names features simultaneously carry out metric transformation, calculate and track the Euclidean distance of target afterwards, find optimal result；
Step 106：Whether the target that judgement has been tracked has the target that can be matched；If no no, into step 107
Then enter step 109；
Step 107：Judge whether to meet establishment fresh target condition；If not meeting, into step 109, if meeting, into step
Rapid 108, enter back into step 109；
Step 108：Create fresh target；
Step 109：The target that renewal is being tracked；
Step 110：First visual angle is processed；
According to one embodiment of present invention, just the processing method at visual angle is in the step 110：First visual angle concentrates on figure
The marginal portion of picture, first determines whether its edge which side is in, and the relative position for tracking target is then found according to its position
Put Color Names features, then carry out metric transformation, calculate Euclidean distance, less than threshold value then be same target, continue with
Track, untill the too small disappearance of this target.
Step 111：Judge whether rectangle frame extracts completely, be detection next frame figure into step 112 if extracting completely
Picture, otherwise, return to step 103 has continued to determine whether two target crossing instances.
According to another aspect of the invention, it is proposed that a kind for the treatment of strategy of target transient loss, specific steps such as Fig. 2 institutes
Show：
Step 201：Input allows the maximum frame number N for losing；
Step 202：Choose two field picture；
Step 203：Whether the target that judgement is being tracked finds similar purpose in this frame；If so, into step 209,
Otherwise enter step 204；
Step 204：Whether the target that judgement is being tracked is to lose for the first time；If so, then entering step 206；Otherwise hold
Row step 205；
Step 205：Lose frame number do Jia 1 process；Into step 207；
Step 206：Frame number is lost to count again；Into step 207；
Step 207：Whether judgement allows the maximum frame number N for losing not less than loss frame number, if so, into step 208, it is no
Then enter step 209；
Step 208：Stop tracking this target；Into step 209；
Step 209：The target that renewal is being tracked.
The above method sets a maximum frame number for allowing to lose, and cannot be matched when using Color Names features
During the target for tracking, then temporarily its status indication to lose, when the target as matching is found again, then according to preceding
The target location for detecting twice afterwards, calculates position mean, adds among the target for tracking；If continuous lose
Frame number be more than or equal to allow lose max-thresholds N, then stop tracking target.
With reference to the explanation of the invention and practice that disclose here, other embodiment of the invention is for those skilled in the art
All will be readily apparent and understand.Illustrate and embodiment is to be considered only as exemplary, true scope of the invention and purport are equal
It is defined in the claims.
Claims (4)
1. the sane motion target tracking method of a kind of high speed, it is characterised in that including step：
A) motion detection is carried out to two field picture；
B) the big target of motion and the Small object of motion are rejected；
C) determine whether that two targets intersect, if not having, into step d, otherwise, according to the target for tracking motion
Target Segmentation is opened, and enters back into step d；
D) according to the size for tracking target, Color Names are extracted in the nearest position of distance original target in intersection region
Feature simultaneously carries out metric transformation, calculates and track the Euclidean distance of target afterwards, finds optimal result；
E) judge whether the target for having tracked has the target that can be matched；If no, into step f, otherwise into step g；
F) judge whether to meet establishment fresh target condition；If not meeting, into step g, if meeting, fresh target is created, then enter
Enter step g；
G) target for tracking is updated；
H) first visual angle is processed；
I) judge whether rectangle frame extracts completely, if extracting completely, into next two field picture, otherwise, return to step c continues to judge
Whether two target crossing instances are had.
2. method according to claim 1, it is characterised in that：Judge the side whether two targets intersect in the step c)
Method is：Two or more target ranges of previous frame image are close, and have close trend.
3. method according to claim 1, it is characterised in that：Just the processing method at visual angle is in the step h)：Just regard
Angle concentrates on the marginal portion of image, first determines whether its edge which side is in, and is then found according to its position and tracks mesh
Target relative position Color Names features, then carry out metric transformation, calculate Euclidean distance, are then same less than threshold value
Target, continues to track, untill the too small disappearance of this target.
4. a kind of target transient loss processing method of method for tracking target suitable for described in claim 1, including step：
Step 201：Input allows the maximum frame number N for losing；
Step 202：Choose two field picture；
Step 203：Whether the target that judgement is being tracked finds similar purpose in this frame；If so, into step 209, otherwise
Into step 204；
Step 204：Whether the target that judgement is being tracked is to lose for the first time；If so, then entering step 206；Otherwise perform step
Rapid 205；
Step 205：Lose frame number do Jia 1 process；Into step 207；
Step 206：Frame number is lost to count again；Into step 207；
Step 207：Whether judgement allows the maximum frame number N for losing not less than frame number is lost, if so, into step 208, otherwise entering
Enter step 209；
Step 208：Stop tracking this target；Into step 209；
Step 209：The target that renewal is being tracked.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611095524.9A CN106780555A (en) | 2016-12-01 | 2016-12-01 | A kind of sane motion target tracking method of high speed |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611095524.9A CN106780555A (en) | 2016-12-01 | 2016-12-01 | A kind of sane motion target tracking method of high speed |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106780555A true CN106780555A (en) | 2017-05-31 |
Family
ID=58883516
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611095524.9A Pending CN106780555A (en) | 2016-12-01 | 2016-12-01 | A kind of sane motion target tracking method of high speed |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106780555A (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101783015A (en) * | 2009-01-19 | 2010-07-21 | 北京中星微电子有限公司 | Equipment and method for tracking video |
| CN102156989A (en) * | 2011-02-25 | 2011-08-17 | 崔志明 | Vehicle blocking detection and segmentation method in video frame |
| CN102881022A (en) * | 2012-07-20 | 2013-01-16 | 西安电子科技大学 | Concealed-target tracking method based on on-line learning |
| CN103456030A (en) * | 2013-09-08 | 2013-12-18 | 西安电子科技大学 | Target tracking method based on scattering descriptor |
| CN103729861A (en) * | 2014-01-03 | 2014-04-16 | 天津大学 | Multiple object tracking method |
| CN104517292A (en) * | 2014-12-25 | 2015-04-15 | 杭州电子科技大学 | Multi-camera high-density crowd partitioning method based on planar homography matrix restraint |
| US20150286638A1 (en) * | 2012-11-09 | 2015-10-08 | Orbeus, Inc. | System, method and apparatus for scene recognition |
| CN105261035A (en) * | 2015-09-15 | 2016-01-20 | 杭州中威电子股份有限公司 | Method and device for tracking moving objects on highway |
| CN105654512A (en) * | 2015-12-29 | 2016-06-08 | 深圳羚羊微服机器人科技有限公司 | Target tracking method and device |
| US20160300118A1 (en) * | 2015-04-07 | 2016-10-13 | Xerox Corporation | Discriminative embedding of local color names for object retrieval and classification |
-
2016
- 2016-12-01 CN CN201611095524.9A patent/CN106780555A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101783015A (en) * | 2009-01-19 | 2010-07-21 | 北京中星微电子有限公司 | Equipment and method for tracking video |
| CN102156989A (en) * | 2011-02-25 | 2011-08-17 | 崔志明 | Vehicle blocking detection and segmentation method in video frame |
| CN102881022A (en) * | 2012-07-20 | 2013-01-16 | 西安电子科技大学 | Concealed-target tracking method based on on-line learning |
| US20150286638A1 (en) * | 2012-11-09 | 2015-10-08 | Orbeus, Inc. | System, method and apparatus for scene recognition |
| CN103456030A (en) * | 2013-09-08 | 2013-12-18 | 西安电子科技大学 | Target tracking method based on scattering descriptor |
| CN103729861A (en) * | 2014-01-03 | 2014-04-16 | 天津大学 | Multiple object tracking method |
| CN104517292A (en) * | 2014-12-25 | 2015-04-15 | 杭州电子科技大学 | Multi-camera high-density crowd partitioning method based on planar homography matrix restraint |
| US20160300118A1 (en) * | 2015-04-07 | 2016-10-13 | Xerox Corporation | Discriminative embedding of local color names for object retrieval and classification |
| CN105261035A (en) * | 2015-09-15 | 2016-01-20 | 杭州中威电子股份有限公司 | Method and device for tracking moving objects on highway |
| CN105654512A (en) * | 2015-12-29 | 2016-06-08 | 深圳羚羊微服机器人科技有限公司 | Target tracking method and device |
Non-Patent Citations (3)
| Title |
|---|
| ERIC P. XING 等: "Distance Metric Learning, with Application to Clustering with Side-Information", 《PROCEEDING NIPS’02 PROCESSDINGS OF THE 15TH INTERNATIONAL CONFERENCE ON NEURAL INFORMATION PROCESSING SYSTEM》 * |
| MARTIN DANELLJAN 等: "Adaptive Color Attributes for Real-Time Visual Tracking", 《IEEE CONFERENCE ON COMPUTER VISION & PATTERN RECOGNITION,2014》 * |
| YANG YANG 等: "Salient Color Names for Person Re-Identication", 《COMPUTER VISION-ECCV 2014: 13TH EUROPEAN CONFERENCE. ZURICH. SWITZERLAND.》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106407917B (en) | Retinal Vessel Extraction Method and System Based on Dynamic Scale Allocation | |
| CN102722891B (en) | Method for detecting image significance | |
| Prati et al. | Shadow detection algorithms for traffic flow analysis: a comparative study | |
| CN102800095B (en) | Lens boundary detection method | |
| CN105205489B (en) | Detection method of license plate based on color and vein analyzer and machine learning | |
| CN112818862A (en) | Face tampering detection method and system based on multi-source clues and mixed attention | |
| CN103955949B (en) | Moving target detecting method based on Mean-shift algorithm | |
| CN101854467B (en) | Method for adaptively detecting and eliminating shadow in video segmentation | |
| CN102324025A (en) | Face detection and tracking method based on Gaussian skin color model and feature analysis | |
| CN102799863A (en) | Method for detecting group crowd abnormal behaviors in video monitoring | |
| CN108960185A (en) | Vehicle target detection method and system based on YOLOv2 | |
| CN103034870B (en) | The boats and ships method for quickly identifying of feature based | |
| CN105005766A (en) | Vehicle body color identification method | |
| CN105205804A (en) | Caryon-cytolymph separation method and apparatus of white blood cells in blood cell image, and classification method and apparatus of white blood cells in blood cell image | |
| CN102930334A (en) | Video recognition counter for body silhouette | |
| CN103942539A (en) | Method for accurately and efficiently extracting human head ellipse and detecting shielded human face | |
| CN111553214A (en) | Method and system for detecting smoking behavior of drivers | |
| US12437404B1 (en) | Machine vision-based method for rapidly detecting scar contour | |
| CN105631456B (en) | A kind of leucocyte method for extracting region based on particle group optimizing ITTI model | |
| CN101610412B (en) | A visual tracking method based on multi-cue fusion | |
| CN106446832B (en) | Video-based pedestrian real-time detection method | |
| CN105701474A (en) | Method for identifying video smog in combination with color and appearance characteristics | |
| Wang et al. | Deep learning-based human activity analysis for aerial images | |
| CN106780555A (en) | A kind of sane motion target tracking method of high speed | |
| CN104899898B (en) | Pavement detection method based on multidimensional information probabilistic model |
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 |
Application publication date: 20170531 |
|
| RJ01 | Rejection of invention patent application after publication |