CN104732558B - moving object detection device - Google Patents
moving object detection device Download PDFInfo
- Publication number
- CN104732558B CN104732558B CN201310716765.0A CN201310716765A CN104732558B CN 104732558 B CN104732558 B CN 104732558B CN 201310716765 A CN201310716765 A CN 201310716765A CN 104732558 B CN104732558 B CN 104732558B
- Authority
- CN
- China
- Prior art keywords
- pixel
- module
- background
- dynamic image
- processing module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The present invention provides a kind of moving object detection device, and it includes photographing module, memory module, reference record module and processing module, and the photographing module is to obtain dynamic image;The memory module connects the photographing module, and to store the dynamic image in a cycle period, latest tendency image covers dynamic image at first;The reference record module connects the memory module, and the reference record module obtains pixel average, pixel undulating value and the pixel prospect number of dynamic image in a cycle period;The processing module connects the reference record module, and the processing module judges whether background model needs reconstruction according to pixel undulating value and pixel prospect number, if desired rebuilds, using pixel average as background model rebuild when initial Gaussian distribution average.The present invention can save memory space and call number, so as to ensure the real-time of system.
Description
【Technical field】
It is particularly a kind of to avoid because memory capacity influences the motion of system the present invention relates to a kind of moving object detection device
Object detecting device.
【Background technology】
In recent years, it has been subjected to increasingly as Detection for Moving Target more crucial in intelligent Video Surveillance Technology
The concern of many people, existing video monitoring system can be only done simple tracking and record, the detection of video content and analyzing and processing
Work, which is still relied on, to be accomplished manually.The delayed development and application for seriously having hindered video monitoring system of Video Analysis Technology.
As most basic process part, the moving object detection of video image in intelligent monitoring, video compress, lead automatically
There is extensive prospect in terms of boat, man-machine interaction, virtual reality.Moving object detection in Video Supervision Technique refers to from even
Target is extracted in continuous video flowing, region or color of target etc. is typically to determine.However, due to weather and the change of illumination
The presence of the reason such as change, interference, the shadow of moving target and the camera motion of background clutter motion, to moving object detection
Correctness bring great challenge.Can correctly it be tracked because the correct detection of moving target affects moving target with segmentation
With classification, therefore moving object detection turn into video monitoring system research in an important problem.Its basic task is
Detect movable information from image sequence, simplify image processing process, obtain required motion vector, so as to recognize with
Track object.
Background model method is that the static background statistical model of monitoring scene is set up in the case of existing moving target, is passed through
Moving target is distinguished in the contrast of present frame and background model.Background modeling is by analyzing scene image background pixel value
Variation Features, for target monitoring scene, with mathematical modeling come simulation context image, set up the description mould of scene image background
Type.During target detection, matched with each frame pixel value in picture frame to be checked with background descriptive model, the pixel mark of matching
Background is designated as, unmatched pixel is labeled as target.At present, realize in the moving object detection in outdoor scene, it is the most frequently used
The probabilistic model for describing background dot distribution is mixed Gauss model.Mixed Gauss model is excellent in the complicated dynamic background of description
Point be it will be apparent that be still updated using background area, and to the foreground area in current frame image without updating,
Although so making background model not contain prospect, a problem is brought, diplopia is easily exactly produced into detection process, from
And influence the correct extraction of moving target.
In order to overcome easy this shortcoming of generation diplopia in traditional update method, it is necessary to introduce institute between successive image frame
Comprising information, but multiple image is stored in internal memory to cause overhead huge, and called from external memory
Two field picture in the past then needs the I/O interfaces of Reusability at a slow speed, influences the real-time of system.
【The content of the invention】
It is a primary object of the present invention to provide a kind of avoid because memory capacity influences the moving object detection device of system.
The present invention provides a kind of moving object detection device, it include photographing module, memory module, reference record module and
Processing module, the photographing module is to obtain dynamic image;The memory module connects the photographing module, and to store
Dynamic image in one cycle period, latest tendency image covers dynamic image at first;The reference record module connects institute
Memory module is stated, and the reference record module obtains the pixel average of dynamic image, pixel undulating value in a cycle period
With pixel prospect number;The processing module connects the reference record module, and the processing module according to pixel undulating value and
Pixel prospect number judge background model whether need reconstruction, if desired rebuild, using pixel average as background model rebuild when
Initial Gaussian is distributed average.
Especially, the processing module is according to the first comparative result and pixel of pixel undulating value/between T and first threshold
Prospect number/the second comparative result between T and Second Threshold judges whether background model needs reconstruction, when pixel undulating value/T is big
When first threshold and pixel prospect number/T are less than Second Threshold, background model needs to rebuild.
Especially, the first threshold is 0.85.
Especially, the Second Threshold is 0.2.
Especially, the reference record module obtains the number of times that pixel is background dot, and processing module is according to pixel
The number of times of background dot dynamically updates the method for reconstructing of background model.
Especially, when the number of times that pixel is background dot is 0, processing module selects adjacent interframe method to rebuild background mould
Type.
Especially, when the number of times that pixel is background dot is three threshold values of 0-, processing module select adjacent interframe method with
The mixed Gaussian of background subtraction fusion rebuilds background model.
Especially, when the number of times that pixel is background dot is not less than three threshold values, processing module selection background subtraction
Rebuild background model.
Compared with prior art, the dynamic image in one cycle period of present invention storage, can save memory space, and
Background model reconstruction is carried out using the pixel average of dynamic image in a cycle period, the call number of picture frame can be reduced,
So as to ensure the real-time of system.
【Brief description of the drawings】
Fig. 1 is the functional-block diagram of moving object detection device of the present invention.
【Embodiment】
Refer to shown in Fig. 1, the present invention provides a kind of moving object detection device, it includes photographing module 10, storage mould
Block 20, reference record module 30 and processing module 40, the photographing module 10 is to obtain dynamic image;The memory module 20
The photographing module 10 is connected, and to store the dynamic image in a cycle period T, the covering of latest tendency image is moved at first
State image;The reference record module 30 connects the memory module 20, and the reference record module 30 obtains a circulation
The pixel average of dynamic image, pixel undulating value and pixel prospect number in cycle;The processing module 40 connects the parameter note
Module 30 is recorded, and the processing module 40 judges whether background model needs reconstruction according to pixel undulating value and pixel prospect number,
If desired rebuild, using pixel average as background model rebuild when initial Gaussian be distributed average.
In the present embodiment, the processing module 40 is compared knot according to pixel undulating value/first between T and first threshold
Fruit and pixel prospect number/the second comparative result between T and Second Threshold judge whether background model needs reconstruction;Cycle period
In T, coordinate position (x, the y of the pixel in dynamic image)Represent;Pixel average is represented with u (T, x, y);Pixel prospect
Number is represented with f (T, x, y).Pixel average u (T, x, y) is cycle period T internal coordinates position (x, a y)The average of pixel;
Pixel undulating value d (T, x, y) and pixel prospect number f (T, x, y) is used to judge cycle period T internal coordinates position (x, a y)Picture
Whether vegetarian refreshments occurs prospect background conversion, if judging to convert, indicates that coordinate position (x, y)Original back of the body of pixel
Scape model is no longer applicable, it is necessary to coordinate position (x, y)The background model of pixel is rebuild.Judge whether to need to rebuild
Method it is as follows:When f (T, x, y)/T is less than Second Threshold more than first threshold, and d (T, x, y)/T, illustrate coordinate position
(x,y)The background model of pixel needs to rebuild.Wherein, f (T, x, y)/T indicates coordinate position (x, y)Pixel is upper one
The number of times of prospect is judged as in the individual cycle, sets first threshold to be difficult too small, background model available letter during due to rebuilding
Breath is less, sets the value of first threshold to be preferably not less than in 0.7, the present embodiment, first threshold takes 0.85.D (T, x, y)/T shows
One cycle period T internal coordinates position (x, y)The severe degree of pixel change, if setting, Second Threshold is excessive, and having can
It can will move slower object to rebuild as background, in the present embodiment, Second Threshold takes 0.2.
The reference record module 30 obtains the number of times that pixel is background dot, and processing module is background dot according to pixel
Number of times dynamically update background model method for reconstructing.
When background constructing and reconstruction, the information of former frame is only existed in background model, pixel is the number of times of background dot
For 0, the difference that should be more laid particular emphasis on during moving target between analysis consecutive frame, i.e. neighbor frame difference method are judged whether, because
This, processing module 40 selects adjacent interframe method to rebuild background model.
Background model rebuild during, pixel be background dot number of times be three threshold values of 0- when, processing module 40 is selected
The mixed Gaussian that adjacent interframe method is merged with background subtraction rebuilds background model.
When Background Modeling after a period of time, pixel is that the number of times of background dot is not less than the 3rd threshold value, the back of the body
Scape model is more stable, and the information of consecutive frame is updated for background model and small obtain also is wanted in the contribution of target detection relatively
It is many, in order to reduce the noise that background perturbation is brought, it should the difference laid particular emphasis between analysis image and background model, i.e. background
Subtraction, therefore, the selection background subtraction of processing module 40 rebuild background model.
Dynamic image in the present invention one cycle period of storage, can save memory space, and utilize a cycle period
The pixel average of interior dynamic image carries out background model reconstruction, the call number of picture frame can be reduced, so as to ensure system
Real-time.
The foregoing is only a specific embodiment of the invention, but protection scope of the present invention is not limited thereto, any
Those familiar with the art the invention discloses technical scope in, change or replacement can be readily occurred in, should all be contained
Cover within protection scope of the present invention.Therefore, protection scope of the present invention should be defined by scope of the claims.
Claims (4)
1. a kind of moving object detection device, it is characterised in that including:
Photographing module, it is to obtain dynamic image;
Memory module, it connects the photographing module, and to store the dynamic image in a cycle period T, latest tendency
Image covers dynamic image at first;
Reference record module, it connects the memory module, and the reference record module obtains each pixel of the dynamic image
Pixel average, pixel undulating value, the number of times as background and the number of times as prospect o'clock in a cycle period;
Processing module, it connects the reference record module, and the processing module is according to pixel undulating value and pixel prospect number
Judge whether background model needs reconstruction, if desired rebuild, using pixel average as background model rebuild when initial Gaussian point
Cloth average, and be that the number of times of background dot dynamically updates the method for reconstructing of background model according to pixel.
2. moving object detection device according to claim 1, it is characterised in that:When the number of times that pixel is background dot is
When 0, processing module selects adjacent interframe method to rebuild background model.
3. moving object detection device according to claim 1, it is characterised in that:When the number of times that pixel is background dot is
During three threshold values of 0-, the mixed Gaussian that processing module selects adjacent interframe method to be merged with background subtraction rebuilds background model.
4. moving object detection device according to claim 1, it is characterised in that:When the number of times that pixel is background dot is
During not less than three threshold values, processing module selection background subtraction rebuilds background model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310716765.0A CN104732558B (en) | 2013-12-20 | 2013-12-20 | moving object detection device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310716765.0A CN104732558B (en) | 2013-12-20 | 2013-12-20 | moving object detection device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104732558A CN104732558A (en) | 2015-06-24 |
CN104732558B true CN104732558B (en) | 2017-10-27 |
Family
ID=53456426
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310716765.0A Active CN104732558B (en) | 2013-12-20 | 2013-12-20 | moving object detection device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104732558B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108629254B (en) * | 2017-03-24 | 2020-11-27 | 杭州海康威视数字技术股份有限公司 | Moving target detection method and device |
CN108805878A (en) * | 2018-05-22 | 2018-11-13 | 深圳腾视科技有限公司 | A kind of foreground object detection solution based on computer vision |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102903124A (en) * | 2012-09-13 | 2013-01-30 | 苏州大学 | Moving object detection method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007089763A (en) * | 2005-09-28 | 2007-04-12 | Toshiba Corp | Radiolucent image processor, radiolucent image processing method and radiolucent image processing program |
-
2013
- 2013-12-20 CN CN201310716765.0A patent/CN104732558B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102903124A (en) * | 2012-09-13 | 2013-01-30 | 苏州大学 | Moving object detection method |
Non-Patent Citations (2)
Title |
---|
Yunchu Zhang等.An adaptive mixture Gaussian background model with online background reconstruction and adjustable.《Industrial Technology, 2005. ICIT 2005. IEEE International Conference on》.2005,第23-27页. * |
梁英宏等.一种快速的视频序列运动分割方法.《计算机工程》.2008,第34卷(第12期),第197-199页. * |
Also Published As
Publication number | Publication date |
---|---|
CN104732558A (en) | 2015-06-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Oh et al. | Fast video object segmentation by reference-guided mask propagation | |
CN102567727B (en) | Method and device for replacing background target | |
CN110837778A (en) | Traffic police command gesture recognition method based on skeleton joint point sequence | |
CN101470809B (en) | Moving object detection method based on expansion mixed gauss model | |
CN109670405B (en) | Complex background pedestrian detection method based on deep learning | |
CN106296725A (en) | Moving target detects and tracking and object detecting device in real time | |
CN102307274A (en) | Motion detection method based on edge detection and frame difference | |
CN103105924B (en) | Man-machine interaction method and device | |
CN105809716B (en) | Foreground extraction method integrating superpixel and three-dimensional self-organizing background subtraction method | |
CN103258332A (en) | Moving object detection method resisting illumination variation | |
Song et al. | Design of control system based on hand gesture recognition | |
CN106228109A (en) | A kind of action identification method based on skeleton motion track | |
CN108256567B (en) | Target identification method and system based on deep learning | |
CN108710879B (en) | Pedestrian candidate region generation method based on grid clustering algorithm | |
CN103810718A (en) | Method and device for detection of violently moving target | |
CN114170570A (en) | Pedestrian detection method and system suitable for crowded scene | |
WO2022001961A1 (en) | Detection method, detection device and detection system for moving target thrown from height | |
CN103456012B (en) | Based on visual human hand detecting and tracking method and the system of maximum stable area of curvature | |
CN106127810B (en) | A kind of the recording and broadcasting system image tracking method and device of the light stream of video macro block angle point | |
CN104732558B (en) | moving object detection device | |
Zhang et al. | An optical flow based moving objects detection algorithm for the UAV | |
CN114463800A (en) | Multi-scale feature fusion face detection and segmentation method based on generalized intersection-parallel ratio | |
CN107274382A (en) | A kind of state identification method, device and the electronic equipment of hard pressing plate | |
CN113822161B (en) | Dynamic hand detection method and device for fusion of skin color and three-level background updating model | |
CN110858392A (en) | Monitoring target positioning method based on fusion background model |
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 |