CN103473547A - Vehicle target recognizing algorithm used for intelligent traffic detecting system - Google Patents
Vehicle target recognizing algorithm used for intelligent traffic detecting system Download PDFInfo
- Publication number
- CN103473547A CN103473547A CN2013104331028A CN201310433102A CN103473547A CN 103473547 A CN103473547 A CN 103473547A CN 2013104331028 A CN2013104331028 A CN 2013104331028A CN 201310433102 A CN201310433102 A CN 201310433102A CN 103473547 A CN103473547 A CN 103473547A
- Authority
- CN
- China
- Prior art keywords
- background
- algorithm
- binary image
- image
- difference
- 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 relates to a target recognizing method based on video detection, and discloses a vehicle target recognizing algorithm used for an intelligent traffic detecting system. The algorithm includes the first step of background model building, the second step of background updating, the third step of background difference, the fourth step of multiframe difference, the fifth step of target outline obtaining by means of combination of the multiframe difference and the background difference, wherein the fifth step includes the steps that (E1) continuous multiframe images are collected and stored, (E2) a binary image D1 is obtained through a frame difference algorithm of the first frame image and the second frame image, a binary image D2 is obtained through the frame difference algorithm of the second frame image and the third frame image, and in the same way, a binary image D3 and a binary image D4 are obtained, (E3) an and operation is carried out on the D1 and the D2 so as to obtain a binary image D5, an and operation is carried out on the D3 and the D4 so as to obtain a binary image D6, and an and operation is carried out on the D5 and the D6 so as to obtain a final binary image D7, and (E4) an and operation is carried out on a binary image obtained from the binary image D7 and the background difference in the third step so that the outline of a vehicle target can be recognized.
Description
Technical field
The present invention relates to a kind of method that detects the identification object based on video, specifically, related to a kind of vehicle target thing recognizer be applied in the Intelligent traffic video detection system.
Background technology
It is the key component in intelligent transportation system that the vehicle target thing extracts with recognizer, and wherein the identification of vehicle target thing has several different methods, comprises frame difference method, background subtraction point-score, edge detection method and algorithm of region growing etc.Vehicle identification is the important component part of intelligent transportation system, increasingly mature due to the development of video equipment and high-performance computer hardware and video processing technique, vehicle recongnition technique based on video becomes the focus of intelligent transportation field research, and obtains applying more and more widely.
Foreground extraction and recognition methods are particularly important in vehicle identification.Foreground extracting method commonly used mainly contains: edge detection method, method of difference, algorithm of region growing etc.Wherein, method of difference comprises again frame difference method and background subtraction method.Edge detection method: because the pixel of target and background inside has lower Grad, and its border has higher Grad, therefore can carry out by the statistics to Grad in image the detection at edge, i.e. edge detection method.Its shortcoming is: edge detection algorithm commonly used exist computing consuming time many, efficiency is low, in addition, for some, is not target clearly, edge detection algorithm detects not out.Secondly, noise can cause and detect pseudo-edge.
Frame difference method is the target identification method often used at present, subtracted each other by adjacent two two field picture respective pixel the value obtained, with the threshold value set in advance, compare, if be greater than threshold value, just there is moving target in explanation, and carries out binary conversion treatment, by the zone marker that is defined as moving target, it is foreground pixel, otherwise, just there is no moving target, and it be labeled as to background pixel.The shortcoming of its existence is: the slow or static prospect of moving can be missed; Thereby likely in target internal, produce cavity and can't intactly extract moving target; There is certain hysteresis quality, produce " hangover " phenomenon, and the faster this situation of the speed of target travel will be more serious.
The background subtraction point-score is a kind of effective moving object detection algorithm, basic thought is the background image that first obtains the ROI zone, and preserve, then the background image respective pixel of current frame image and preservation is subtracted to computing, difference is greater than the foreground pixel that is labeled as of setting threshold, otherwise be labeled as background pixel, and it is carried out to binary conversion treatment, just can identify the foreground target in image.Its shortcoming is the algorithm of background extraction template and the automatic renewal relative complex of background; And hysteresis quality is more serious, when environmental change and background does not upgrade in time, just there will be the target of obtaining inaccurate.
Algorithm of region growing: the basic thought of algorithm of region growing is that the set of pixels with similar quality is formed to zone altogether.At first a Seed Points is looked in the zone that will cut apart each, then will in neighborhood around sub pixel, with the pixel of sub pixel by same or similar character, merge in the zone at sub pixel place.These new pixels are used as to new Seed Points and proceed said process, until the pixel do not satisfied condition.Its shortcoming is: the algorithm of region growing complexity is relatively high, calculates the time used longer, is not suitable for the system high to requirement of real-time; If the beginning of growth criterion and termination condition are selected with great difficulty not cause that program enters endless loop, and manually appointment of growing point, robustness is poor.
The Chinese invention patent that publication No. is CN102509101 " background update method in traffic video monitoring and vehicle target are put forward the power method " is only applicable to desirable environment, topmost drawback is:<1 > if rain, snow, the exceedingly odious weather of wind, be difficult to obtain a desirable background, the undesirable direct accuracy that the vehicle target thing extracts that affects of background; The shade of the non-vehicle target thing that<2 > the sunlight oblique fire occurs, very easily cause erroneous judgement and fail to judge;<3 > ideally can obtain good background by day, but can not obtain desirable background equally at night, directly affect the accuracy that the vehicle target thing extracts.
Summary of the invention
The present invention is directed to the prior art deficiency, propose a kind of algorithm that can accurately identify the vehicle target thing, even also can identify accurately the vehicle target thing in daytime and evening under rugged environment.
To achieve these goals, the technical solution adopted in the present invention is:
A kind of thing of the vehicle target for intelligent transportation detection system recognizer comprises following step:
Steps A, read continuous multiple image, appoint and get a kind of background modeling algorithm, set up background model;
Step B, context update;
Step C, background difference;
Step D, multi-frame difference;
Step e, the identification of vehicle target thing; The profile that adopts multi-frame difference and background subtraction to divide the method combined to obtain object, its step is as follows:
E1, at first collect continuous multiple image and preserve;
E2, the first two field picture and the second two field picture carry out the frame difference algorithm and obtain binary image D1, and the second two field picture and the 3rd two field picture carry out the frame difference algorithm and obtain binary image D2, by that analogy, obtain respectively binary image D3 and D4;
E3, D1 and D2 carry out obtaining binary image D5 with computing, and D3 and D4 carry out obtaining D6 with computing, and D5 and D6 carry out obtaining final binary image D7 with computing;
E4, the binary image D7 obtained by E3, and in step C, background subtraction divides the binary image obtained to carry out and computing, the complete profile that identifies the vehicle target thing.
Described vehicle target thing recognizer, in step B, adopt the context update algorithm to obtain desirable background; Described context update algorithm comprises following step:
Step B1, calibrate in image the coordinate that needs the sensitizing range detected;
Step B2, the continuous multiple image to obtaining, the method combined with frame difference algorithm and background difference algorithm, the initial background of the sensitizing range of demarcating in obtaining step B1;
Step B3, carry out the context update judgement;
Step B4, context update;
Described vehicle target thing recognizer, in step B3, the step of carrying out the context update judgement is as follows:
C1, obtain distinguishing the threshold value of present frame and background frames with function nerve chain network algorithm;
C2, obtain the number of pixels of certain gray level with the grey level histogram algorithm;
C3, current frame image and background image are carried out to difference, if the current pixel difference result is greater than the threshold value of setting, thinks so and have moving target, be labeled as foreground pixel, otherwise, be labeled as background pixel; The part that then will be labeled as background pixel is substituted in pixel corresponding in background template goes, to realize the real-time update of background.
The useful good effect of invention:
1, vehicle target thing recognizer of the present invention, night and daytime the vehicle target thing discrimination can reach GB, efficiently solve in Intelligent traffic video analytic system algorithm to environment according to lazyness.The method that adopts background subtraction criterion and the poor criterion of frame to combine, undertaken and computing by the binary image that obtains in the background subtraction sub-module and the binary image of described multi-frame difference, even under exceedingly odious weather, also can accurately identify the vehicle target thing.
2, vehicle target thing recognizer of the present invention, have the identification of intellectuality, object precisely, the recognition efficiency advantages of higher.1) even rain, snow, the exceedingly odious weather of wind also can be easy to obtain desirable background, and can carry out context update fast and effectively according to weather condition; 2) can effectively avoid shade occurring because sunlight causes object, cause the problem of vehicle target thing identification error.
3, vehicle target thing recognizer of the present invention, the method of utilizing the poor criterion of frame and background difference criterion to combine is learnt and the real-time update background, judge whether to carry out context update according to update condition, in the time of can effectively avoiding context update, static target is updated as a setting, not only efficiently solve the traditional intelligence video analytic system to the height of environment according to lazyness; And by the method that adopts five frame differences and foreground image to do " with computing ", realize efficiently solving the identification of vehicle target thing the object identified and have empty defect.
The accompanying drawing explanation
Fig. 1: vehicle target thing recognizer process flow diagram of the present invention.
Embodiment
Below by embodiment, technical scheme of the present invention is described in further detail.
embodiment 1
Referring to Fig. 1.The present invention, for the vehicle target thing recognizer of intelligent transportation detection system, comprising: steps A, read continuous multiple image, appoint and get a kind of background modeling algorithm, set up background model;
Step B, context update;
Step C, background difference;
Step D, multi-frame difference;
Step e, the identification of vehicle target thing: the profile that adopts multi-frame difference and background subtraction to divide the method combined to obtain object, its step is as follows:
E1, at first collect continuous multiple image and preserve;
E2, the first two field picture and the second two field picture carry out the frame difference algorithm and obtain binary image D1, and the second two field picture and the 3rd two field picture carry out the frame difference algorithm and obtain binary image D2, by that analogy, obtain respectively binary image D3 and D4;
E3, D1 and D2 carry out obtaining binary image D5 with computing, and D3 and D4 carry out obtaining D6 with computing, and D5 and D6 carry out obtaining final binary image D7 with computing;
E4, the binary image D7 obtained by E3, and in step C, background subtraction divides the binary image obtained to carry out and computing, the complete profile that identifies the vehicle target thing.
embodiment 2
Vehicle target for the intelligent transportation detection system thing recognizer of the present embodiment, as different from Example 1: in step B, specifically adopt following context update algorithm to obtain desirable background:
Step B1, calibrate in image the coordinate that needs the sensitizing range detected;
Step B2, the continuous multiple image to obtaining, the method combined with frame difference algorithm and background difference algorithm, the initial background of the sensitizing range of demarcating in obtaining step B1;
Step B3, carry out the context update judgement;
Step B4, context update.
embodiment 3
Vehicle target for the intelligent transportation detection system thing recognizer of the present embodiment, difference from Example 2 is:
In step B3, the step of carrying out the context update judgement is as follows:
C1, obtain distinguishing the threshold value of present frame and background frames with function nerve chain network algorithm;
C2, obtain the number of pixels of certain gray level with the grey level histogram algorithm;
C3, current frame image and background image are carried out to difference, if the current pixel difference result is greater than the threshold value of setting, thinks so and have moving target, be labeled as foreground pixel, otherwise, be labeled as background pixel; The part that then will be labeled as background pixel is substituted in pixel corresponding in background template goes, to realize the real-time update of background.
Claims (3)
1. the thing of the vehicle target for an intelligent transportation detection system recognizer comprises following step:
Steps A, read continuous multiple image, appoint and get a kind of background modeling algorithm, set up background model;
Step B, context update;
Step C, background difference;
Step D, multi-frame difference;
Step e, the identification of vehicle target thing;
It is characterized in that, in step e, the profile that adopts multi-frame difference and background subtraction to divide the method combined to obtain object, its step is as follows:
E1, at first collect continuous multiple image and preserve;
E2, the first two field picture and the second two field picture carry out the frame difference algorithm and obtain binary image D1, and the second two field picture and the 3rd two field picture carry out the frame difference algorithm and obtain binary image D2, by that analogy, obtain respectively binary image D3 and D4;
E3, D1 and D2 carry out obtaining binary image D5 with computing, and D3 and D4 carry out obtaining D6 with computing, and D5 and D6 carry out obtaining final binary image D7 with computing;
E4, the binary image D7 obtained by E3, and in step C, background subtraction divides the binary image obtained to carry out and computing, the complete profile that identifies the vehicle target thing.
2. vehicle target thing recognizer according to claim 1, is characterized in that: in step B, adopt the context update algorithm to obtain desirable background; Described context update algorithm comprises following step:
Step B1, calibrate in image the coordinate that needs the sensitizing range detected;
Step B2, the continuous multiple image to obtaining, the method combined with frame difference algorithm and background difference algorithm, the initial background of the sensitizing range of demarcating in obtaining step B1;
Step B3, carry out the context update judgement;
Step B4, context update.
3. vehicle target thing recognizer according to claim 2, it is characterized in that: in step B3, the step of carrying out the context update judgement is as follows:
C1, obtain distinguishing the threshold value of present frame and background frames with function nerve chain network algorithm;
C2, obtain the number of pixels of certain gray level with the grey level histogram algorithm;
C3, current frame image and background image are carried out to difference, if the current pixel difference result is greater than the threshold value of setting, thinks so and have moving target, be labeled as foreground pixel, otherwise, be labeled as background pixel; The part that then will be labeled as background pixel is substituted in pixel corresponding in background template goes, to realize the real-time update of background.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013104331028A CN103473547A (en) | 2013-09-23 | 2013-09-23 | Vehicle target recognizing algorithm used for intelligent traffic detecting system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013104331028A CN103473547A (en) | 2013-09-23 | 2013-09-23 | Vehicle target recognizing algorithm used for intelligent traffic detecting system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103473547A true CN103473547A (en) | 2013-12-25 |
Family
ID=49798391
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013104331028A Pending CN103473547A (en) | 2013-09-23 | 2013-09-23 | Vehicle target recognizing algorithm used for intelligent traffic detecting system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103473547A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103914701A (en) * | 2014-03-20 | 2014-07-09 | 燕山大学 | Night vehicle detection method based on image |
| CN103984926A (en) * | 2014-05-15 | 2014-08-13 | 江苏科大汇峰科技有限公司 | Distributed moving object detection method based on MapReduce calculation model |
| CN104599511A (en) * | 2015-02-06 | 2015-05-06 | 中国石油大学(华东) | Traffic flow detection method based on background modeling |
| CN104867332A (en) * | 2015-05-26 | 2015-08-26 | 南京通用电器有限公司 | Interval-frame difference method based detection method of driving vehicle in front lane line |
| CN105447890A (en) * | 2015-12-08 | 2016-03-30 | 南京航空航天大学 | Motion vehicle detection method resisting light effect |
| CN106845346A (en) * | 2016-12-16 | 2017-06-13 | 北京无线电计量测试研究所 | A kind of image detecting method for airfield runway foreign bodies detection |
| CN110263693A (en) * | 2019-06-13 | 2019-09-20 | 苏州元联传感技术有限公司 | In conjunction with the traffic detection recognition method of inter-frame difference and Bayes classifier |
| CN115273565A (en) * | 2022-06-24 | 2022-11-01 | 苏州数智源信息技术有限公司 | Airplane apron early warning method, device and terminal based on AI big data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040151342A1 (en) * | 2003-01-30 | 2004-08-05 | Venetianer Peter L. | Video scene background maintenance using change detection and classification |
| CN102184552A (en) * | 2011-05-11 | 2011-09-14 | 上海理工大学 | Moving target detecting method based on differential fusion and image edge information |
| CN103258432A (en) * | 2013-04-19 | 2013-08-21 | 西安交通大学 | Traffic accident automatic identification processing method and system based on videos |
-
2013
- 2013-09-23 CN CN2013104331028A patent/CN103473547A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040151342A1 (en) * | 2003-01-30 | 2004-08-05 | Venetianer Peter L. | Video scene background maintenance using change detection and classification |
| CN102184552A (en) * | 2011-05-11 | 2011-09-14 | 上海理工大学 | Moving target detecting method based on differential fusion and image edge information |
| CN103258432A (en) * | 2013-04-19 | 2013-08-21 | 西安交通大学 | Traffic accident automatic identification processing method and system based on videos |
Non-Patent Citations (3)
| Title |
|---|
| 任朝栋 等: "快速消除车辆阴影的多阈值图像分割法", 《应用光学》 * |
| 刘向军: "基于局部特征匹配的目标跟踪研究", 《中国优秀硕士学位论文全文数据库信息科技辑2012年》 * |
| 李红艳: "交通监控系统中的运动目标检测与分类", 《中国优秀硕士学位论文全文数据库信息科技辑2012年》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103914701A (en) * | 2014-03-20 | 2014-07-09 | 燕山大学 | Night vehicle detection method based on image |
| CN103914701B (en) * | 2014-03-20 | 2017-10-27 | 燕山大学 | A kind of vehicle detection at night method based on image |
| CN103984926A (en) * | 2014-05-15 | 2014-08-13 | 江苏科大汇峰科技有限公司 | Distributed moving object detection method based on MapReduce calculation model |
| CN104599511A (en) * | 2015-02-06 | 2015-05-06 | 中国石油大学(华东) | Traffic flow detection method based on background modeling |
| CN104867332A (en) * | 2015-05-26 | 2015-08-26 | 南京通用电器有限公司 | Interval-frame difference method based detection method of driving vehicle in front lane line |
| CN104867332B (en) * | 2015-05-26 | 2017-06-30 | 南京通用电器有限公司 | Based on the detection method every driving vehicle in the front lane line of frame difference method |
| CN105447890A (en) * | 2015-12-08 | 2016-03-30 | 南京航空航天大学 | Motion vehicle detection method resisting light effect |
| CN106845346A (en) * | 2016-12-16 | 2017-06-13 | 北京无线电计量测试研究所 | A kind of image detecting method for airfield runway foreign bodies detection |
| CN110263693A (en) * | 2019-06-13 | 2019-09-20 | 苏州元联传感技术有限公司 | In conjunction with the traffic detection recognition method of inter-frame difference and Bayes classifier |
| CN115273565A (en) * | 2022-06-24 | 2022-11-01 | 苏州数智源信息技术有限公司 | Airplane apron early warning method, device and terminal based on AI big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103473547A (en) | Vehicle target recognizing algorithm used for intelligent traffic detecting system | |
| CN104616290A (en) | Target detection algorithm in combination of statistical matrix model and adaptive threshold | |
| CN103456028B (en) | A kind of moving target detecting method | |
| CN108022249B (en) | Automatic extraction method for target region of interest of remote sensing video satellite moving vehicle | |
| WO2022027931A1 (en) | Video image-based foreground detection method for vehicle in motion | |
| CN102663362B (en) | Moving target detection method based on gray features | |
| CN102332167A (en) | Target detection method for vehicles and pedestrians in intelligent traffic monitoring | |
| CN101872546A (en) | Video-based method for rapidly detecting transit vehicles | |
| CN102568005A (en) | Moving object detection method based on Gaussian mixture model | |
| CN104063885A (en) | Improved movement target detecting and tracking method | |
| CN108052917A (en) | A kind of method of the architecture against regulations automatic identification found based on new and old Temporal variation | |
| CN104952256A (en) | Video information based method for detecting vehicles at intersection | |
| CN109711256B (en) | Low-altitude complex background unmanned aerial vehicle target detection method | |
| CN103986906A (en) | Door opening and closing detection method based on monitoring videos | |
| CN103258332A (en) | Moving object detection method resisting illumination variation | |
| CN104899881A (en) | Shadow detection method for moving vehicle in video image | |
| EP2813973A1 (en) | Method and system for processing video image | |
| CN103679704A (en) | Video motion shadow detecting method based on lighting compensation | |
| CN103793921B (en) | Moving object extraction method and moving object extraction device | |
| CN103020980A (en) | Moving target detection method based on improved double-layer code book model | |
| CN104182976A (en) | Field moving object fining extraction method | |
| CN108010050B (en) | Foreground detection method based on adaptive background updating and selective background updating | |
| CN103065145A (en) | Vehicle movement shadow eliminating method | |
| CN105205833A (en) | Moving object detection method and device based on space-time background model | |
| CN105469394B (en) | A kind of Intelligent target tracking based on complex environment |
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 |