CN105208398B - A kind of method for obtaining the real-time Background of road - Google Patents
A kind of method for obtaining the real-time Background of road Download PDFInfo
- Publication number
- CN105208398B CN105208398B CN201510608645.8A CN201510608645A CN105208398B CN 105208398 B CN105208398 B CN 105208398B CN 201510608645 A CN201510608645 A CN 201510608645A CN 105208398 B CN105208398 B CN 105208398B
- Authority
- CN
- China
- Prior art keywords
- frame
- cluster
- road
- grid
- maximum
- 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.)
- Expired - Fee Related
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of methods for obtaining the real-time Background of road, use Grid Clustering quick obtaining road Background, road is divided into many grids first, secondly these grids are carried out with the statistics of H components, then cluster that different video frame is polymerized to is determined according to H component curve figures, finally therefrom extraction feature frame, finally therefrom extraction feature frame.This method has the advantages that complexity is low, efficient, exploitativeness is strong.
Description
Technical field
The invention belongs to digital picture processing and cluster field, especially suitable for road video extraction Background.Road is carried on the back
The modeling of scape figure is a kind of effective ways of video compress transmission process, and in particular to a kind of new Grid Clustering extracts Background
The characteristic frame of piece.
Background technology
Recently, background modeling plays an increasingly important role in high efficiency monitor video coding.Meanwhile many practicalities regard
Frequency coding application also proposed background modeling some specific requirements, such as low carrying cost, low computation complexity.
Existing background modeling method can substantially be classified as 2 classes：Parametric method, as gauss hybrid models -1, Gauss mix
Molding type -2, gauss hybrid models -3 and nonparametric technique include Bayesian model, Density Estimator, time medium filtering,
Value drift etc..
The drawbacks such as these methods have mathematical formulae more, and coding is complicated, operational efficiency is low, there is an urgent need for using it is new, simple easily
Row and efficiently method are for background modeling.
It is existing with obtaining the relevant side of road background diagram technology by being found to the retrieval of existing patent and the relevant technologies
Method and system include：
⑴Low-complexity and high-efficiency background modeling for
Surveillance video coding, 2012IEEE International Conference on Visual
Communication and Image Processing,San Jose,USA,pp.1-6,11/2012
Propose (SWRA) method based on a branch and weight.First by each pixel in SWRA in training frames
Position be divided into some time, then calculate its corresponding average value and weight.Later, weighted mean procedure is used to reduce
Influence foreground pixel, and obtain modeling result.
⑵A Fuzzy Background Modeling Approach for Motion Detection in
Dynamic Backgrounds, Multimedia and Signal Processing Volume 346of the series
Communications in Computer and Information Science pp 177-185
A kind of method is proposed, in -2 model of Gaussian Mixture, is come using fuzzy logic system recurrence sef-adapting filter
The update weight of calculating, and finally obtain road Background.As a result it proves using the fuzzy method compared with conventional method,
With larger advantage.
⑶Difference of Gaussian Edge-Texture Based Background Modeling for
Dynamic Traffic Conditions, Advances in Visual Computing Volume 5358of the
series Lecture Notes in Computer Science pp 406-417
It proposes based on Gauss Edge texture, obtains road background and the method for probe vehicle.This method passes through pixel
Point and its edge, non-edge pixels point establish contact, have very strong learning performance, can be good at visiting road prospect
It surveys and sorts out.
It can be seen that above method, which is all based on Gauss model, carries out mileage chart processing, have formula complicated, coding is realized
Difficulty is big, does not have the shortcomings that real-time, is not suitable for road monitoring field higher to requirement of real-time now.
In existing patent still not explicitly for cluster obtain road Background method, so it is proposed that based on net
The method that lattice cluster obtains road Background has preferable research significance and application value.
Invention content
In view of the deficiency of existing program set forth above, the present invention is intended to provide efficiently, simple method, and be allowed to overcome
The disadvantage mentioned above of the prior art.
To achieve these goals, the considerations of of the invention is：
The road of normally travel all has vehicle above each frame picture.But the concept of grid is used, each frame
Video is divided into tiny grid.Video pictures in each grid are made color histogram by us with the H components in HSV,
By the maximum value of all H components of the grid in a period of time, curve graph is made.When carless in grid, H components
Maximum value curve is basicly stable.When have automobile by when, particularly vehicle color is with the presence of having arrived Background color
When larger difference, the maximum value of H components will appear great variety.Vehicle color and road are just the same or same
Color automobile series by probability be substantially not present.
Its specific processing includes following means：
A kind of method for obtaining the real-time Background of road, it is first using the method for Grid Clustering quick obtaining road Background
Road is first divided into many grids, secondly these grids are carried out with the statistics of H components, is then determined according to H component curve figures
The cluster that different video frame is polymerized to, finally therefrom extraction feature frame, includes following processing means：
(1) it is frame by the road video extraction of 30 seconds, video 30 frame per second, 900 frame altogether；
(2) grid interception is done to every frame to divide, and 100 lattice is divided into per frame picture, respectively with the two-dimensional matrix of A-J and 1-10
It represents；
(3) to H components in each grid computing HSV of each frame, to the H component statistical maximum values of each grid；By 900
The maximum value of H components in each grid of each frame of frame is depicted as curve graph；The fluctuation of curve illustrates main in this period
There is offset in color, has vehicle to pass through the region；
(4) with the method extraction characteristic frame of cluster：
A. it is calculated since grid A1, is cluster by point cluster of all deviation threshold ranges within 10, the maximum
Cluster is exactly the big cluster of road surface Background；
B. the point cluster of all identical values is cluster in the maximum cluster obtained during a is walked, finds out the cluster of number maximum, takes
Its longest continuity point, and extract characteristic frame；Characteristic frame pkExtraction model be：
Wherein, pkFor the characteristic frame extracted, pnFor each frame in maximum cluster, n is the number of frame in maximum cluster, and i is most
The number that first frame starts in big cluster, j are the number of last frame in maximum cluster.
(5) characteristic frame got is replaced into the original frame of the grid, and returns to the extraction that (4) step starts next grid
Characteristic frame until J10 processing completions, obtains the real-time Background of whole picture road；
(6) it returns to (1) step and starts next round calculating.
In actual implementation, the division methods of the grid of each frame, can also be according to practical need except being divided into as 100 especially
Determine the grid number divided.
The present invention is directed to the modeling problem of road Background, specifically proposes a kind of feasibility height, and simple, real-time is very
Strong Grid Clustering obtains road background drawing method.
Description of the drawings is as follows：
Fig. 1 is one first frame segmentation figure of video.
Fig. 2 is the H component curve figures in b1 regions.
Fig. 3 is the H component curve figures in b5 regions.
Fig. 4 is the 1st frame picture.
Fig. 5 is the 900th frame picture.
Fig. 6 is road background composite diagram.
Fig. 7 is the flow chart of the method for the present invention.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Because of the interference of extraneous sunlight, cloud layer, leaf, we are can not to obtain accurate curvilinear motion figure.Therefore we
The point changed in threshold range is found out in the graph, and is gathered for cluster.It is considered that here it is road Backgrounds
Cluster.Again in the cluster, continually look for having what stable state changed, have the frame of maximum eigenvalue, and draw it out, as the grid
Background.Mesh generation is as shown in Figure 1, it is specifically included：
(1) it is frame by the road video extraction of 30 seconds.If video 30 frame per second, 900 frame altogether；
(2) do grid interception to every frame to divide, per frame picture, we are divided into 100 lattice, respectively with the two dimension of A-J and 1-10
Matrix represents, such as figure one；
(3) to H components in each grid computing HSV of each frame, to the H component statistical maximum values of each grid.By 900
The maximum value of H components in each grid of each frame of frame is depicted as curve graph.The fluctuation of curve illustrates main in this period
There is offset in color, has vehicle to pass through the region.
(4) with the method extraction characteristic frame of cluster：
Characteristic frame pkExtraction formula：
p:Selected characteristic frame；
i:In (4)-b steps, selected initial position；
j:In (4)-b steps, selected end position.
A. it is calculated since grid A1, it, in this way can be with by point cluster of all deviation threshold ranges within 10 for cluster
It filters out because of the offset of H values caused by leaf swing, sunlight scattering, cloud layer variation；Because when the area road free time
It occupies the majority, it is believed that this maximum cluster is exactly the big cluster of road surface Background；
B. the maximum cluster obtained during a is walked, continues to handle it.It is cluster the point cluster of identical values all in cluster,
Find out the cluster of number maximum.Can find out in this way can most represent the frame of road background under current environment；
C. in the cluster obtained again in b steps, longest continuity point is taken, and extract characteristic frame.By the step for, can be with
Further confirm that the moment, which stablizes the most in ectocine, road is in idle state.We recognize
The real-time Background in the region of present road can be represented for this feature frame, therefore this feature frame can be used to replace originally
Frame.If Fig. 2, Fig. 3 are change curve of the H components maximum value of drafting in 30 seconds.
(5) characteristic frame got is replaced to original characteristic frame, and returns to the extraction spy that (4) step starts next grid
Frame is levied, until J10 processing is completed；
(6) it returns to (1) step and starts next round calculating, as described in Figure 7 step.
From the picture of first frame and the 900th frame it can be found that picture has larger shade variation, such as Fig. 4, Fig. 5, figure
Described in 6.Movement due to cloud layer, the variation of light are can be seen that from Fig. 6 composite diagrams, different grid charts is that have shade
Difference.It can be concluded that from experiment, the real-time of road Background is replaced, to reducing Internet video flow,
It is of practical significance.
From above experiment is discussed, it can know that this method has following clear advantage：
1st, this method has that extent feasible is very high, the simple advantage of coding；
2nd, the Background that this method obtains has stronger real-time；
3rd, the characteristic frame of this method extraction can react many variations such as cloud layer movement, light variation, and to road in time
The shaking of leaf has very strong anti-interference on both sides of the road.
Claims (1)
- A kind of 1. method for obtaining the real-time Background of road, which is characterized in that use Grid Clustering quick obtaining road Background Method, road is divided into many grids first, secondly to these grids carry out H components statistics, then according to H components song Line chart determines the cluster that different video frame is polymerized to, finally therefrom extraction feature frame, includes following processing step：(1) it is frame by the road video extraction of 30 seconds, video 30 frame per second, 900 frame altogether；(2) grid interception is done to every frame to divide, and is divided into 100 lattice per frame picture, is represented respectively with the two-dimensional matrix of A-J and 1-10；(3) to H components in each grid computing HSV of each frame, to the H component statistical maximum values of each grid；By 900 frames The maximum value of H components in each each grid of frame is depicted as curve graph；The fluctuation of curve illustrates that master color occurs in this 30 seconds Offset, has the vehicle to pass through measured zone；(4) with the method extraction characteristic frame of cluster：A. it is calculated since grid A1, is cluster by point cluster of all deviation threshold ranges within 10, the cluster of the maximum is just It is the big cluster of road surface Background；B. the point cluster of all identical values is cluster in the maximum cluster obtained during a is walked, finds out the cluster of number maximum, takes it most Long continuity point, and extract characteristic frame；Characteristic frame pkExtraction model be：Wherein, pkFor the characteristic frame extracted, pnFor each frame in maximum cluster, n is the number of frame in maximum cluster, and i is maximum cluster The number that middle first frame starts, j are the number of last frame in maximum cluster；(5) characteristic frame got is replaced into the original frame of the grid, and returns to the extraction feature that (4) step starts next grid Frame until J10 processing completions, obtains the real-time Background of whole picture road；(6) it returns to (1) step and starts next round calculating.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510608645.8A CN105208398B (en) | 2015-09-22 | 2015-09-22 | A kind of method for obtaining the real-time Background of road |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510608645.8A CN105208398B (en) | 2015-09-22 | 2015-09-22 | A kind of method for obtaining the real-time Background of road |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105208398A CN105208398A (en) | 2015-12-30 |
| CN105208398B true CN105208398B (en) | 2018-06-19 |
Family
ID=54955785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510608645.8A Expired - Fee Related CN105208398B (en) | 2015-09-22 | 2015-09-22 | A kind of method for obtaining the real-time Background of road |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105208398B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0993443A (en) * | 1995-05-16 | 1997-04-04 | Sanyo Electric Co Ltd | Color-monochromatic image conversion method and edge position detection method for object to be inspected |
| CN101533515A (en) * | 2009-04-13 | 2009-09-16 | 浙江大学 | Background modeling method based on block facing video monitoring |
| CN101834981A (en) * | 2010-05-04 | 2010-09-15 | 崔志明 | Video background extracting method based on online cluster |
| CN102722720A (en) * | 2012-05-25 | 2012-10-10 | 苏州大学 | Video background extraction method based on hue-saturation-value (HSV) space on-line clustering |
| CN103985114A (en) * | 2014-03-21 | 2014-08-13 | 南京大学 | Surveillance video person foreground segmentation and classification method |
| JP2015121901A (en) * | 2013-12-20 | 2015-07-02 | 日本放送協会 | Video area division device and video area division program |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101215987B1 (en) * | 2008-12-22 | 2012-12-28 | 한국전자통신연구원 | Apparatus for separating foreground from back ground and method thereof |
-
2015
- 2015-09-22 CN CN201510608645.8A patent/CN105208398B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0993443A (en) * | 1995-05-16 | 1997-04-04 | Sanyo Electric Co Ltd | Color-monochromatic image conversion method and edge position detection method for object to be inspected |
| CN101533515A (en) * | 2009-04-13 | 2009-09-16 | 浙江大学 | Background modeling method based on block facing video monitoring |
| CN101834981A (en) * | 2010-05-04 | 2010-09-15 | 崔志明 | Video background extracting method based on online cluster |
| CN102722720A (en) * | 2012-05-25 | 2012-10-10 | 苏州大学 | Video background extraction method based on hue-saturation-value (HSV) space on-line clustering |
| JP2015121901A (en) * | 2013-12-20 | 2015-07-02 | 日本放送協会 | Video area division device and video area division program |
| CN103985114A (en) * | 2014-03-21 | 2014-08-13 | 南京大学 | Surveillance video person foreground segmentation and classification method |
Non-Patent Citations (7)
| Title |
|---|
| a fuzzy backgrournd modeling approach for motion detection in dynamic backgournds;YING DING ET.AL;《Mutimedia and signal processing 》;20090918;第177至185页 * |
| forground object detection in complex scences using cluster color;chung chi lin et.al;《2014 8th international conference on innovative moblie and internet services in ubiquitous computing》;20140602;第529页至532页 * |
| low-complexity and high-efficiency background mdodeling for surveillance video coding;Xinguo Zhang et.al;《visual communications and image processing,2012 IEEE》;20121130;第1至第6页 * |
| 基于HSV颜色空间的视频车辆检测与跟踪算法研究;赵作升;《中国优秀硕士全文数据库》;20090215;I138-478 * |
| 基于网格模型的运动估计特征点的提取;杨坤;《中国优秀硕士学位论文全文数据库》;20081115;I138-731 * |
| 应用色彩空间聚类方法道路建模;向宸薇;《中国图像图形学报》;20130816;第18卷(第8期);0976页至0981页 * |
| 改进的道路背景提取与更新算法;李洁等;《视频应用与工程》;20130612;第37卷(第11期);194页至1977页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105208398A (en) | 2015-12-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108230264B (en) | Single image defogging method based on ResNet neural network | |
| CN100578563C (en) | Vehicle count method based on video image | |
| CN101957997B (en) | Regional average value kernel density estimation-based moving target detecting method in dynamic scene | |
| CN113592026B (en) | Binocular vision stereo matching method based on cavity volume and cascade cost volume | |
| CN103077530A (en) | Moving object detection method based on improved mixing gauss and image cutting | |
| CN111402126B (en) | Video super-resolution method and system based on blocking | |
| CN102044073B (en) | Method and system for judging crowd density in image | |
| CN103871076A (en) | Moving object extraction method based on optical flow method and superpixel division | |
| CN105631898A (en) | Infrared motion object detection method based on spatio-temporal saliency fusion | |
| CN106991686A (en) | A kind of level set contour tracing method based on super-pixel optical flow field | |
| CN105303581A (en) | Adaptive parameter moving target detection method | |
| CN103514608A (en) | Movement target detection and extraction method based on movement attention fusion model | |
| CN104616026A (en) | Monitor scene type identification method for intelligent video monitor | |
| CN110111267A (en) | A kind of single image based on optimization algorithm combination residual error network removes rain method | |
| CN105046670A (en) | Image rain removal method and system | |
| CN1266656C (en) | Intelligent alarming treatment method of video frequency monitoring system | |
| CN114155210A (en) | Crowd counting method based on attention mechanism and standardized dense void space multi-scale fusion network | |
| CN106778789B (en) | A kind of fast target extracting method in multi-view image | |
| CN110503049B (en) | Satellite video vehicle number estimation method based on generation countermeasure network | |
| CN111931722A (en) | Correlated filtering tracking method combining color ratio characteristics | |
| CN108508425A (en) | Foreground target detection method based on neighborhood characteristics under a kind of radar near-earth ambient noise | |
| CN105208398B (en) | A kind of method for obtaining the real-time Background of road | |
| CN117350927A (en) | Image rain removing method based on deep learning | |
| CN104268889A (en) | Vehicle tracking method based on automatic characteristic weight correction | |
| CN104063879A (en) | Pedestrian flow estimation method based on flux and shielding coefficient |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180619 Termination date: 20210922 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |