CN103473792A - Method for detecting moving target - Google Patents
Method for detecting moving target Download PDFInfo
- Publication number
- CN103473792A CN103473792A CN2013104116854A CN201310411685A CN103473792A CN 103473792 A CN103473792 A CN 103473792A CN 2013104116854 A CN2013104116854 A CN 2013104116854A CN 201310411685 A CN201310411685 A CN 201310411685A CN 103473792 A CN103473792 A CN 103473792A
- Authority
- CN
- China
- Prior art keywords
- frame
- moving target
- moving object
- gmm
- moving
- 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
Images
Abstract
The invention discloses a method for detecting a moving target. The method comprises the following steps of (a), reading in a video frame, (b), preprocessing the video frame and enabling the video frame to become a gray level image frame, (c), judging whether the image frame is a first frame or not, if yes, building a GMM, and if not, carrying out the step (d), (d), processing the image frame, and obtaining characteristics of the moving target; (e), fusing the characteristics and obtaining image information of the moving object, (f), computing pixel points of the moving object to obtain the mass center of the moving object, and storing the position, in the frame, of the moving object, (g), comparing the mass center with a set threshold value, judging whether the mass center changes or not between two frames, if the mass center does not change, carrying out the step (i), and if the mass center changes, carrying out the step (h), (h), carrying out morphological processing and updating the GMM, (i) carrying out the morphological processing and not updating the GMM, (j) outputting the detection result of the current frame. The method can solve the problem that detection is not accurate when the moving object keeps static for a long time or when the moving object moves slowly, and is particularly suitable for detecting moving objects under the sea.
Description
Technical field
The present invention relates to video image and process and identify, particularly relate to a kind of moving target detecting method.
Background technology
Detection for the thalassogenic movement target in video, the prior art general process is all by ripe modeling technique, the seabed video is carried out to the behaviouristics that modeling analysis is not considered moving target, simple set up background model from background and carry out difference mobilizing exercises target, and the renewal of background does not rely on the behavior of moving target.
For example, the moving object detection algorithmic procedure that the people such as Yin Mengni propose at " Harbin University of Commerce's journal " in 2012 is: improved gauss hybrid models moving object detection algorithm, its whole flow process may be summarized to be five parts: image importation, Gaussian Mixture moving object detection part, SUSAN rim detection part, Fusion Features part and output as a result, wherein in the second part Gaussian Mixture moving object detection part not to the behavior analysis of moving target.
Consider the uncertainty of moving target, if target is static a period of time somewhere, existing Target Modeling technology will be upgraded gauss hybrid models, this target is processed as background, this defect will directly cause in the loss of moving target and differential process occurring " cavity " phenomenon, and the background model of now upgrading will be affected, the detection of moving target after being unfavorable for, also can be greatly affected the behavior analysis of moving target.
Summary of the invention
The object of the invention is to overcome the deficiencies in the prior art, provide that a kind of detection efficiency is high, the moving target detecting method of accurate positioning, solve the long-time transfixion of moving object or mobilely cause detecting inaccurate problem in situation slowly.
For achieving the above object, the present invention is by the following technical solutions:
A kind of moving target detecting method comprises the following steps:
A) read in frame of video;
B) by the frame of video pre-service, be grey-level image frame; .
C) judge that whether picture frame is the first frame, if it is sets up gauss hybrid models, if not turning to steps d).
D) picture frame is processed, obtained the moving target feature;
E) carry out Fusion Features, obtain the moving object image information;
F) calculate the moving object pixel and obtain its barycenter, and preserve its position in frame.
G) compare with setting threshold, judge whether barycenter changes between two continuous frames, if do not change, turn to step I), if change, turn to step h);
H) carry out the morphology processing and upgrade gauss hybrid models;
I) carry out the morphology processing but do not upgrade gauss hybrid models;
J) output present frame testing result.
Can further adopt following technical schemes:
Steps d) comprising:
GMM background subtraction portions, carry out the background difference by the GMM background model to having set up, to extract the moving target unique point;
The rim detection part, obtain the object information in image by rim detection, to extract the moving target unique point.
Described GMM background subtraction portions comprises gauss hybrid models coupling and gauss hybrid models background extracting.
Step e), in, described Fusion Features comprises that the moving target unique point obtained respectively by GMM background difference and rim detection carries out logic and operation.
Described morphology processing comprises at least one in morphological dilations, burn into filtration, refinement and pruning modes.
The moving target detecting method that the present invention proposes, " close-loop feedback theory " in the kybernetics subject is incorporated in target detection, the moving target that monitoring has in real time detected, by result feedback in context update mechanism, obtain effective background model, thereby can not only effectively detect the moving target of " always moving ", can also accurately and reliably detect the moving target of " motionless once in a while ".
In traditional detection moving object method, do not consider the long-time transfixion of moving object or mobile slow this situation, and the present invention calculates the centroid position that obtains moving object by pixel, and the centroid position that calculates two continuous frames changes, whether surpass by judging that this changes the renewal that setting threshold is controlled whole background, like this, with regard to effectively avoided by motionless object be integrated in the background of renewal cause detect inaccurate.
The invention solves the problem of context update in the long-time transfixion of moving object or mobile situation slowly, prevent because the fusion of prospect and background causes the moving object detection and tracking generation of situation to occur missing and can not find.The present invention can be detected long-time transfixion or mobile slow moving object, for the follow-up tracking to this object and analysis provide condition.Compared with prior art, moving target detecting method of the present invention has advantages of that detection efficiency is high, accurate positioning, is particularly useful for the detection of thalassogenic movement object.
The accompanying drawing explanation
The moving target detecting method process flow diagram that Fig. 1 is a kind of preferred embodiment of the present invention.
Embodiment
Below in conjunction with accompanying drawing, embodiments of the invention are elaborated.Should be emphasized that, following explanation is only exemplary, rather than in order to limit the scope of the invention and to apply.
Consult Fig. 1, in some embodiments, moving target detecting method comprises the following steps:
Step a) is read in frame of video, can comprise original video is transformed into to video image one by one.
Step b) by frame of video frame by frame pre-service be gray level image, can comprise pre-service such as picture frame being carried out to smoothing, and transform into gray level image, think that subsequent step prepares.
Whether the picture frame that step c) judgement is read in is the first frame, if it is sets up the GMM(gauss hybrid models), if not, turn to steps d).
Setting up GMM is the gauss hybrid models initialization, can select the random pixel value in video image the first frame to be made as the average of gauss hybrid models, and set weights and the variance of gauss hybrid models.Average and variance can be set as the case may be.In initialization procedure, as far as possible large (as 15) that generally variance arranged, weights are (as 0.001) as small as possible.
Steps d) preferably by following two parts, the picture frame read in is processed, is obtained the moving target feature:
First is GMM background subtraction portions:
GMM background subtraction portions is mainly to carry out the background difference to obtain moving target unique point F1 by the GMM background model to having set up.GMM background subtraction portions can comprise:
I. gauss hybrid models mates
Constantly carry out the foundation of background image since the second picture frame, a frame video image judges whether pixel is consistent with the condition of Model Matching, if meet this pixel and gauss hybrid models, is complementary.
Ii. gauss hybrid models background extracting
Background extracting is the background difference, and by the gauss hybrid models of having set up, " subtraction " obtains foreground image, obtains unique point F1.
Second portion is the rim detection part:
The rim detection part is mainly to carry out rim detection to obtain the object information in image, by rim detection, obtains moving target unique point F2.
In preferred embodiment, by above-mentioned GMM background difference and rim detection, obtain respectively moving target unique point F1 and F2, next by logic and operation, can obtain than only adopting a kind of wherein method to process more effective clear moving object.
Step e) Fusion Features, obtain moving object.
Step f) calculate the moving object pixel and obtain its barycenter, and preserve its position at this frame.
Step g) compare with setting threshold, judge whether barycenter changes between two continuous frames; If do not change, turn to step I), if change, turn to step h).
Step h) carry out the morphology processing and upgrade GMM.
Upgrading GMM, is mainly average/weights of upgrading the front random initializtion/variance, pixel is carried out to the classification of prospect and background, and the average/weights that upgrade here/variance all can be utilized known corresponding update mode.
Step I) only carry out the morphology processing and do not upgrade GMM.
Step j) output present frame testing result.
Morphology is processed can comprise expansion, corrosion treatment etc.Morphology is processed and to be usingd digital morphological and learn and extracts from image as instrument for expression and the useful picture content of description region shape, such as border, skeleton and convex hull.Morphology is processed and also can be comprised morphology filtration, refinement and the pruning etc. for pre-service or aftertreatment.
Above content is in conjunction with concrete preferred implementation further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For the general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, can also make some simple deduction or replace, all should be considered as belonging to protection scope of the present invention.
Claims (5)
1. a moving target detecting method, is characterized in that, comprises the following steps:
A) read in frame of video;
B) by the frame of video pre-service, be grey-level image frame;
C) judge that whether picture frame is the first frame, if it is sets up gauss hybrid models, if not turning to steps d);
D) picture frame is processed, obtained the moving target feature;
E) carry out Fusion Features, obtain the moving object image information;
F) calculate the moving object pixel and obtain its barycenter, and preserve its position in frame;
G) compare with setting threshold, judge whether barycenter changes between two continuous frames, if do not change, turn to step I), if change, turn to step h);
H) carry out the morphology processing and upgrade gauss hybrid models;
I) carry out the morphology processing but do not upgrade gauss hybrid models;
J) output present frame testing result.
2. moving target detecting method as claimed in claim 1, is characterized in that, steps d) comprising:
GMM background subtraction portions, carry out the background difference by the GMM background model to having set up, to extract the moving target unique point;
The rim detection part, obtain the object information in image by rim detection, to extract the moving target unique point.
3. moving target detecting method as claimed in claim 2, is characterized in that, described GMM background subtraction portions comprises gauss hybrid models coupling and gauss hybrid models background extracting.
4. moving target detecting method as claimed in claim 2 or claim 3, is characterized in that step e) in, described Fusion Features comprises that the moving target unique point obtained respectively by GMM background difference and rim detection carries out logic and operation.
5. as claim 1 or 2 or 3 described moving target detecting methods, it is characterized in that, described morphology processing comprises at least one in morphological dilations, burn into filtration, refinement and pruning modes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013104116854A CN103473792A (en) | 2013-09-11 | 2013-09-11 | Method for detecting moving target |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013104116854A CN103473792A (en) | 2013-09-11 | 2013-09-11 | Method for detecting moving target |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103473792A true CN103473792A (en) | 2013-12-25 |
Family
ID=49798626
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013104116854A Pending CN103473792A (en) | 2013-09-11 | 2013-09-11 | Method for detecting moving target |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103473792A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103942812A (en) * | 2014-03-12 | 2014-07-23 | 华南理工大学 | Moving object detection method based on Gaussian mixture and edge detection |
CN104599290A (en) * | 2015-01-19 | 2015-05-06 | 苏州经贸职业技术学院 | Video sensing node-oriented target detection method |
CN105574891A (en) * | 2015-12-11 | 2016-05-11 | 上海兴芯微电子科技有限公司 | Method and system for detecting moving object in image |
CN106504273A (en) * | 2016-10-28 | 2017-03-15 | 天津大学 | A kind of innovatory algorithm based on GMM moving object detections |
CN108734720A (en) * | 2017-04-25 | 2018-11-02 | 佳能株式会社 | Detector for moving object, its control method and computer readable storage medium |
CN109241896A (en) * | 2018-08-28 | 2019-01-18 | 腾讯科技(深圳)有限公司 | A kind of channel security detection method, device and electronic equipment |
CN109886935A (en) * | 2019-01-28 | 2019-06-14 | 南京威翔科技有限公司 | A kind of road face foreign matter detecting method based on deep learning |
Citations (2)
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 |
CN102110296A (en) * | 2011-02-24 | 2011-06-29 | 上海大学 | Method for tracking moving target in complex scene |
-
2013
- 2013-09-11 CN CN2013104116854A patent/CN103473792A/en active Pending
Patent Citations (2)
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 |
CN102110296A (en) * | 2011-02-24 | 2011-06-29 | 上海大学 | Method for tracking moving target in complex scene |
Non-Patent Citations (3)
Title |
---|
DAR-SHYANG LEE: "Effective Gaussian Mixture Learning for Video Background Subtraction", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
殷梦妮: "基于高斯混合模型的运动目标检测", 《中国优秀硕士学位论文全文数据库 信息科技辑(月刊)》 * |
王欢: "视频图像序列中运动目标检测算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑(月刊)》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103942812A (en) * | 2014-03-12 | 2014-07-23 | 华南理工大学 | Moving object detection method based on Gaussian mixture and edge detection |
CN103942812B (en) * | 2014-03-12 | 2017-02-08 | 华南理工大学 | Moving object detection method based on Gaussian mixture and edge detection |
CN104599290A (en) * | 2015-01-19 | 2015-05-06 | 苏州经贸职业技术学院 | Video sensing node-oriented target detection method |
CN104599290B (en) * | 2015-01-19 | 2017-05-10 | 苏州经贸职业技术学院 | Video sensing node-oriented target detection method |
CN105574891A (en) * | 2015-12-11 | 2016-05-11 | 上海兴芯微电子科技有限公司 | Method and system for detecting moving object in image |
CN105574891B (en) * | 2015-12-11 | 2019-04-09 | 上海兴芯微电子科技有限公司 | The method and system of moving target in detection image |
CN106504273A (en) * | 2016-10-28 | 2017-03-15 | 天津大学 | A kind of innovatory algorithm based on GMM moving object detections |
CN108734720A (en) * | 2017-04-25 | 2018-11-02 | 佳能株式会社 | Detector for moving object, its control method and computer readable storage medium |
CN109241896A (en) * | 2018-08-28 | 2019-01-18 | 腾讯科技(深圳)有限公司 | A kind of channel security detection method, device and electronic equipment |
CN109886935A (en) * | 2019-01-28 | 2019-06-14 | 南京威翔科技有限公司 | A kind of road face foreign matter detecting method based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103473792A (en) | Method for detecting moving target | |
CN101610420B (en) | Automatic white balance method | |
CN101404086B (en) | Target tracking method and device based on video | |
CN103413120B (en) | Tracking based on object globality and locality identification | |
CN103530894B (en) | A kind of video object method for tracing based on multiple dimensioned piece of rarefaction representation and system thereof | |
Zhang et al. | A vehicle detection algorithm based on three-frame differencing and background subtraction | |
CN103778618A (en) | Method for fusing visible image and infrared image | |
CN107301376B (en) | Pedestrian detection method based on deep learning multi-layer stimulation | |
CN102222346A (en) | Vehicle detecting and tracking method | |
CN102063727B (en) | Covariance matching-based active contour tracking method | |
CN105046683A (en) | Object detection method based on adaptive-parameter-adjustment Gaussian mixture model | |
Xu et al. | Moving object detection based on improved three frame difference and background subtraction | |
CN106504273B (en) | Improved method based on GMM moving object detection | |
CN104680521A (en) | Improved background modeling and foreground detecting method | |
CN105654518B (en) | A kind of trace template adaptive approach | |
CN104598935A (en) | Incremental subspace learning algorithm-based face recognition method | |
CN108010050B (en) | Foreground detection method based on adaptive background updating and selective background updating | |
CN103456012A (en) | Visual human hand detection tracking method and system based on maximum stable curvature region | |
CN108765451A (en) | A kind of movement of traffic object detection method of adaptive RTS threshold adjustment | |
CN105975911A (en) | Energy perception motion significance target detection algorithm based on filter | |
CN102800105A (en) | Target detection method based on motion vector | |
CN104517300A (en) | Vision judgment tracking method based on statistical characteristic | |
CN104240268B (en) | A kind of pedestrian tracting method based on manifold learning and rarefaction representation | |
CN116091551A (en) | Target retrieval tracking method and system based on multi-mode fusion | |
Du et al. | Multi-angle face detection based on improved RFCN algorithm using multi-scale training |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20131225 |
|
RJ01 | Rejection of invention patent application after publication |