CN105205833B - A kind of moving target detecting method and device based on time-and-space background model - Google Patents

A kind of moving target detecting method and device based on time-and-space background model Download PDF

Info

Publication number
CN105205833B
CN105205833B CN201510586019.3A CN201510586019A CN105205833B CN 105205833 B CN105205833 B CN 105205833B CN 201510586019 A CN201510586019 A CN 201510586019A CN 105205833 B CN105205833 B CN 105205833B
Authority
CN
China
Prior art keywords
background
module
target
time
background model
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
Application number
CN201510586019.3A
Other languages
Chinese (zh)
Other versions
CN105205833A (en
Inventor
石旭刚
张水发
刘嘉
杜雅慧
汤泽胜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
OB TELECOM ELECTRONICS CO Ltd
Original Assignee
OB TELECOM ELECTRONICS CO Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by OB TELECOM ELECTRONICS CO Ltd filed Critical OB TELECOM ELECTRONICS CO Ltd
Priority to CN201510586019.3A priority Critical patent/CN105205833B/en
Publication of CN105205833A publication Critical patent/CN105205833A/en
Application granted granted Critical
Publication of CN105205833B publication Critical patent/CN105205833B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

The present invention provides a kind of moving target detecting method based on time-and-space background model, including 1：Collection image modeling obtains time background model, and 2：Initial motion target, 3：Obtain the background area after initial motion target growth；4：Obtain space background model；5：Judge whether to meet space background model profile, 6：Renewal time background model.The present invention also provides a kind of moving object detection device based on time-and-space background model, including image capture module, time background modeling module, moving object detection module, space background modeling module, moving target confirm module and background model update module.The present invention proposes the concept of space background modeling, think the background of foreground object and surrounding has certain difference in the features such as color, if initial motion target is consistent with ambient background in the features such as color, then it is assumed that is flase drop, so as to improve the accuracy rate of moving object detection.

Description

A kind of moving target detecting method and device based on time-and-space background model
Technical field
The present invention relates to field of intelligent video surveillance, is examined more particularly to a kind of moving target based on time-and-space background model Survey method and apparatus.
Background technology
The basic task of moving object detection is to extract moving target from image background from sequence image, from And the movable information of target is obtained, and have great importance to work such as follow-up image classification, target followings, can be larger Degree on Simplified analysis work.Due in actual monitored scene, background be frequently not it is totally stationary, but the moment be in In change, such as：The change of weather, the rocking of the small rule of background, the change of illumination, the target gradually incorporated in background Deng so that background modeling turns into an emphasis and difficulties for moving object detection.
Background subtraction receives extensively because speed is fast, the degree of accuracy is high, can extract the reasons such as complete objective contour Concern and application.Wherein, background modeling method is the core of background subtraction detection moving target.At present, typical background Modeling method has：The methods of mixed Gaussian background modeling method, codebook method, nonparametric background modeling method, ViBe, PBAS.This A little traditional background modeling methods are modeled using statistical nature of the pixel in time series, and underuse space letter , therefore, during target detection, often there is more flase drop in breath, and during context update, easily by prospect more Newly into background, and really the speed of context update is not fast enough.
The content of the invention
The technical problems to be solved by the invention are to provide a kind of moving target detecting method based on time-and-space background model, The target detection that can be applied under complex scene, the change of scene can be effectively adapted to.
Technical scheme is used by the present invention solves above-mentioned technical problem：
A kind of moving target detecting method based on time-and-space background model, comprises the following steps：
Step 1：Background modeling, which is carried out, according to some two field pictures collected obtains time background model,
IMAQ is extracted by image capture module from pending monitor video；
Step 2：Present frame and background model are asked poor, after binaryzation, obtain initial motion target, initial motion target is Foreground picture after binaryzation；
Step 3：For each initial motion target, region growing, the background area after being grown are carried out to background area Domain；
The growth refers to from the eight connectivity neighborhood of initial motion target to outgrowth, the gray value phase with moving target Difference within the specific limits be considered homogeneous region, can extend, until that can not extend so that obtain the initial motion target Affiliated background area；
Step 4：Above-mentioned background area is described using mixed Gauss model, obtains space background model；
Step 5：Judge whether initial motion target meets space background model profile, if met, illustrate it is prospect； If do not met, after denoising, final moving target is obtained；
Step 6：Renewal time background model.
Time background model is for obtaining initial motion target, and the effect of renewal time background model is：Eliminate Influence of the illumination gradual change to extraction foreground target.
While using above-mentioned technical proposal, the present invention can also be used or combined using technology further below Scheme：
The modeling process of the step 1 specifically includes following steps：
Step 1.1：Some two field pictures are gathered, are converted to gray level image, it is assumed that in the sequence positioned at the pixel of point (x, y) Gray value is：Y={ y1(x,y),y2(x,y),...,yN(x,y)}；
Step 1.2：Calculate the average and standard deviation of these sampled values (Y value of each point in step 1.1), given threshold T1, if standard deviation is less than threshold value t1, gathered for one kind；If standard deviation is not less than threshold value t1, by these sampled values With two classes are divided into, average and standard deviation are calculated respectively.
The step 2 specifically includes following steps：
Step 2.1：Current frame image is converted into gray level image, grey scale pixel value is individually subtracted corresponding to model Value, it is otherwise prospect for background if corresponding difference is within 3 times of standard deviations；
Step 2.2：By extracting connected region, foreground pixel is sorted out to obtain initial motion target.
Growth in the step 3 refers to open to background area from the eight connectivity neighborhood of each initial motion target and background Begin to grow, obtain the background area after corresponding region growing.
The step 6 specifically includes following steps：
Step 6.1：To motion target area, it is updated using less turnover rate；
Step 6.2：For background area, it is updated using larger turnover rate；
It is above-mentioned it is larger with it is smaller be step 6.1 and the mutual comparison of step 6.2.
Another technical problem to be solved by this invention is to provide a kind of moving target inspection based on time-and-space background model Device is surveyed, the object detecting device includes image capture module, time background modeling module, moving object detection module, sky Between background modeling module, moving target confirm module and background model update module, described image acquisition module, which is used to gather, supervises Control image, the settling time background model that the time background modeling modeling module is used in step 1, the moving object detection Module is used to extract initial moving target from time background model in step 2, and the space background modeling module is used for The corresponding background area of each initial target is calculated, and establishes corresponding space background model, the moving target is true Recognize module to be used to the initial motion target that step 2 is extracted being put into space background model to be verified, obtain final motion Target, the background model update module are used to be updated background and moving target；
Described image acquisition module connects monitoring device and monitoring image therein is acquired, described image collection mould Block is connected to time background modeling module and is sent to acquired image information, and the time background modeling module is connected to The moving object detection module is simultaneously sent to modeling information, and the moving object detection module is connected to space background modeling Module is simultaneously sent to initial motion target information, and the space background modeling module is connected to the moving target and confirms module And space background modeling information and initial motion target information are sent to, the moving target confirms that module is connected to background mould Type update module is simultaneously sent to final moving target information, and the background model update module is according to different turnover rates to fortune Moving-target and background are updated.
The beneficial effects of the invention are as follows：A kind of moving target detecting method and dress based on time-and-space background model of the present invention It is an integrated solution to put, and specifically has following innovation：1st, the concept of space background modeling is proposed, it is believed that preceding scenery The background of body and surrounding has certain difference in the features such as color, if initial motion target in the features such as color with around Background is consistent, then it is assumed that is flase drop, so as to improve the accuracy rate of moving object detection.2nd, in renewal time background model, Background area uses very fast turnover rate, background change is learnt faster into model, and the motion target area after confirming Using slower turnover rate, moving target static for a long time can gradually be learnt into background, improve and stop in short-term again Moving target recall rate.
Brief description of the drawings
Fig. 1 is a kind of moving target detecting method flow chart based on time-and-space background model of the present invention.
Fig. 2 is a kind of moving object detection structure drawing of device based on time-and-space background model of the present invention.
Embodiment
Embodiment 1, a kind of moving target detecting method based on time-and-space background model, referring to the drawings 1.
The moving target detecting method of the present invention specifically includes following steps：
Step 1：Background modeling, which is carried out, according to 3000 two field pictures collected obtains time background model；
1. be converted to background image, it is assumed that the grey scale pixel value positioned at point (x, y) is in the sequence：Y={ y1(x, y),y2(x,y),...,y3000(x,y)}；
2. calculate the mean μ and standard deviation δ of the i.e. above-mentioned grey scale pixel value of these sampled values：
I in formula represents ith sample point,
If δ<20, then the time background model using the Gauss model as the pixel.If δ >=20, K averages are used Clustering procedure, two classes are divided into, average and standard deviation are calculated respectively, as time background model.
Step 2：Present frame and background model are asked poor, after binaryzation, obtain initial motion target；
1. present frame is converted into gray level image, average corresponding to model is individually subtracted in grey scale pixel value, if accordingly Difference is then background within 3 times of standard deviations, is otherwise prospect；
R in formula represents binary image,
2. by extracting connected region, foreground pixel is sorted out to obtain initial motion target.
Step 3：For each initial motion target, region growing, the background area after being grown are carried out to background area Domain；
Refer specifically to, grown since the eight connectivity neighborhood of each initial motion target and background is to background area, obtain phase Background area after the region growing answered；
Step 4：Above-mentioned background area is described using mixed Gauss model, obtains space background model；
Step 5：Whether initial motion target meets space background model profile, i.e. pixel in initial motion target area The difference of gray value and corresponding space background model average, less than 3 times standard deviations, then be background；Otherwise it is prospect；Connect in advance Region, the region that area is less than 20 is removed, obtains final moving target；
Step 6：Use following formula renewal time background model：
μt=(1- ρ) * μt-1+ρ*yt
δt 2=(1- ρ) * δt-1 2+ρ*(yt-μt)T(yt-μt),
In formula, ρ is turnover rate；T is the time, and yt represents the gray value of current pixel, and μ t, δ t represent the t picture respectively The average and standard deviation of plain gray value, μ t-1, δ t-1 represent the average and standard deviation of the t-1 moment grey scale pixel value, T tables respectively Show matrix；
Motion target area, updated using less turnover rate, ρ=0.00001；
Background area, updated using larger turnover rate, ρ=0.0001.
Embodiment 2, a kind of moving object detection device based on time-and-space background model, referring to the drawings 2.
The moving object detection device of the present invention includes image capture module 1, time background modeling module 2, moving target Detection module 3, space background modeling module 4, moving target confirm module 5 and background model update module 6, described image collection Module 1 is used for acquisition monitoring image, and the time background models the settling time background model that modeling module 2 is used in step 1, The moving object detection module 3 is used to extract initial moving target, the space from time background model in step 2 Background modeling module 4 is used to calculate the corresponding background area of each initial target, and establishes corresponding space background mould Type, the moving target confirm that module 5 is used to the initial motion target that step 2 is extracted being put into space background model to be tested Card, obtains final moving target, and the background model update module 6 is used to be updated background and moving target；
Described image acquisition module 1 connects monitoring device and monitoring image therein is acquired, described image collection Module 1 is connected to time background modeling module 2 and is sent to acquired image information, the time background modeling module 2 It is connected to the moving object detection module 3 and is sent to modeling information, the moving object detection module 3 is connected to space Background modeling module 4 is simultaneously sent to initial motion target information, and the space background modeling module 4 is connected to the motion mesh Mark confirms module 5 and is sent to space background modeling information and initial motion target information, and the moving target confirms module 5 It is connected to background model update module 6 and is sent to final moving target information, the background model update module 6 is not according to Same turnover rate is updated to moving target and background.

Claims (1)

  1. A kind of 1. moving target detecting method based on time-and-space background model, it is characterised in that：It the described method comprises the following steps：
    Step 1：Background modeling, which is carried out, according to some two field pictures collected obtains time background model；
    Step 2：Present frame and time background model are asked poor, after binaryzation, obtain initial motion target；
    Step 3：For each initial motion target, region growing, the background area after being grown are carried out to background area；
    Step 4：Above-mentioned background area is described using mixed Gauss model, obtains space background model；
    Step 5：Judge whether initial motion target meets space background model profile, if met, illustrate it is background；If Do not meet, after denoising, obtain final moving target；
    Step 6：Renewal time background model；
    The modeling process of the step 1 specifically includes following steps：
    Step 1.1：Some two field pictures are gathered, are converted to gray level image, the picture of point (x, y) is located in the sequence of the gray level image Plain gray value is：Y={ y1(x,y),y2(x,y),...,yN(x,y)}；
    Step 1.2：The average and standard deviation of these sampled values, given threshold t1 are calculated, will if standard deviation is less than threshold value t1 It gathers for one kind；If standard deviation is not less than threshold value t1, by these sampled values with two classes are divided into, average and standard are calculated respectively Difference；
    The step 2 specifically includes following steps：
    Step 2.1：Current frame image is converted into gray level image, corresponding average is individually subtracted in grey scale pixel value, if phase Difference is answered within 3 times of standard deviations, then is background, is otherwise prospect；
    Step 2.2：By extracting connected region, foreground pixel is sorted out to obtain initial motion target；
    Growth in the step 3 refers to give birth to since the eight connectivity neighborhood of each initial motion target and background is to background area It is long, obtain the background area after corresponding region growing；
    The step 6 specifically includes following steps：
    Step 6.1：To motion target area, it is updated using less turnover rate；
    Step 6.2：For background area, it is updated using larger turnover rate；
    It is above-mentioned it is larger with it is smaller be step 6.1 and the mutual comparison of step 6.2；
    Described moving target detecting method uses object detecting device, the object detecting device include image capture module, when Between background modeling module, moving object detection module, space background modeling module, moving target confirm module and background model more New module, described image acquisition module are used for acquisition monitoring image, the foundation that the time background modeling module is used in step 1 Time background model, the moving object detection module are used to extract initial motion from time background model in step 2 Target, the space background modeling module are used to calculate the corresponding background area of each initial target, and establish corresponding Space background model, the moving target confirms that module is used to the initial motion target that step 2 is extracted being put into space background Verified in model, obtain final moving target, the background model update module is used to enter background and moving target Row renewal；
    Described image acquisition module connects monitoring device and monitoring image therein is acquired, described image acquisition module point Time background modeling module is not connected to and is sent to acquired image information, and the time background modeling module is connected to The moving object detection module is simultaneously sent to modeling information, and the moving object detection module is connected to space background modeling Module is simultaneously sent to initial motion target information, and the space background modeling module is connected to the moving target and confirms module And space background modeling information and initial motion target information are sent to, the moving target confirms that module is connected to background mould Type update module is simultaneously sent to final moving target information, and the background model update module is according to different turnover rates to fortune Moving-target and background are updated.
CN201510586019.3A 2015-09-15 2015-09-15 A kind of moving target detecting method and device based on time-and-space background model Active CN105205833B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510586019.3A CN105205833B (en) 2015-09-15 2015-09-15 A kind of moving target detecting method and device based on time-and-space background model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510586019.3A CN105205833B (en) 2015-09-15 2015-09-15 A kind of moving target detecting method and device based on time-and-space background model

Publications (2)

Publication Number Publication Date
CN105205833A CN105205833A (en) 2015-12-30
CN105205833B true CN105205833B (en) 2018-03-16

Family

ID=54953493

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510586019.3A Active CN105205833B (en) 2015-09-15 2015-09-15 A kind of moving target detecting method and device based on time-and-space background model

Country Status (1)

Country Link
CN (1) CN105205833B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109151376B (en) * 2017-06-27 2021-09-28 南京东大智能化系统有限公司 Intelligent video analysis monitoring method
CN108174055B (en) * 2017-12-29 2020-11-13 广东工业大学 Intelligent monitoring method, system, equipment and storage medium
JP7106282B2 (en) 2018-01-30 2022-07-26 キヤノン株式会社 Image processing device, image processing method and program
CN108921078A (en) * 2018-06-27 2018-11-30 天津大学 Time-and-space background based on video models sport foreground detection algorithm

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007323572A (en) * 2006-06-05 2007-12-13 Nec Corp Object detector, object detection method, and object detection program
CN101325690A (en) * 2007-06-12 2008-12-17 上海正电科技发展有限公司 Method and system for detecting human flow analysis and crowd accumulation process of monitoring video flow
CN101635852A (en) * 2009-08-26 2010-01-27 北京航空航天大学 Method for detecting real-time moving object based on adaptive background modeling
CN102568002A (en) * 2011-12-20 2012-07-11 福建省华大数码科技有限公司 Moving object detection algorithm based on fusion of texture pattern and movement pattern
CN102568005A (en) * 2011-12-28 2012-07-11 江苏大学 Moving object detection method based on Gaussian mixture model
KR20130063963A (en) * 2011-12-07 2013-06-17 한국전자통신연구원 Image processing method for detecting moving object

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007323572A (en) * 2006-06-05 2007-12-13 Nec Corp Object detector, object detection method, and object detection program
CN101325690A (en) * 2007-06-12 2008-12-17 上海正电科技发展有限公司 Method and system for detecting human flow analysis and crowd accumulation process of monitoring video flow
CN101635852A (en) * 2009-08-26 2010-01-27 北京航空航天大学 Method for detecting real-time moving object based on adaptive background modeling
KR20130063963A (en) * 2011-12-07 2013-06-17 한국전자통신연구원 Image processing method for detecting moving object
CN102568002A (en) * 2011-12-20 2012-07-11 福建省华大数码科技有限公司 Moving object detection algorithm based on fusion of texture pattern and movement pattern
CN102568005A (en) * 2011-12-28 2012-07-11 江苏大学 Moving object detection method based on Gaussian mixture model

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"基于时空的混合高斯背景建模的运动目标检测";郭晓 等;《电视技术》;20131231;第37卷(第3期);185-187 *
"基于自适应混合高斯模型的运动目标检测";文如泉 等;《萍乡高等专科学校校报》;20120630;第29卷(第3期);29-32 *

Also Published As

Publication number Publication date
CN105205833A (en) 2015-12-30

Similar Documents

Publication Publication Date Title
CN111784685B (en) Power transmission line defect image identification method based on cloud edge cooperative detection
CN108022249B (en) Automatic extraction method for target region of interest of remote sensing video satellite moving vehicle
Rakibe et al. Background subtraction algorithm based human motion detection
CN110378288A (en) A kind of multistage spatiotemporal motion object detection method based on deep learning
CN102307274A (en) Motion detection method based on edge detection and frame difference
CN105205833B (en) A kind of moving target detecting method and device based on time-and-space background model
CN107194559B (en) Workflow identification method based on three-dimensional convolutional neural network
CN102509085A (en) Pig walking posture identification system and method based on outline invariant moment features
CN103226712B (en) A kind of remnant object detection method based on finite state machine
CN108010047A (en) A kind of moving target detecting method of combination unanimity of samples and local binary patterns
CN109508741A (en) Method based on deep learning screening training set
CN105973904A (en) Edible oil impurity detection method based on image background probability graph
CN105023272A (en) Crop leaf insect pest detection method and system
CN101299274B (en) Detecting method and system for moving fixed target
CN106570885A (en) Background modeling method based on brightness and texture fusion threshold value
CN112489055B (en) Satellite video dynamic vehicle target extraction method fusing brightness-time sequence characteristics
CN117152735A (en) Tomato maturity grading method based on improved yolov5s
CN105405153A (en) Intelligent mobile terminal anti-noise interference motion target extraction method
CN112866654A (en) Intelligent video monitoring system
CN108010050B (en) Foreground detection method based on adaptive background updating and selective background updating
CN109035296A (en) A kind of improved moving objects in video detection method
CN104751487A (en) Method for detecting movement target based on colored RGB three-pane color-change frame difference
CN108734050A (en) Moving object detection and valid frame extracting method in a kind of monitor video
CN107301655B (en) Video moving target detection method based on background modeling
CN109978916A (en) Vibe moving target detecting method based on gray level image characteristic matching

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant