CN105208339A - Accident detection method for recognizing vehicle collision through monitoring videos - Google Patents
Accident detection method for recognizing vehicle collision through monitoring videos Download PDFInfo
- Publication number
- CN105208339A CN105208339A CN201510614372.8A CN201510614372A CN105208339A CN 105208339 A CN105208339 A CN 105208339A CN 201510614372 A CN201510614372 A CN 201510614372A CN 105208339 A CN105208339 A CN 105208339A
- Authority
- CN
- China
- Prior art keywords
- light stream
- accident
- impingement
- collision
- behavior
- 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
Landscapes
- Traffic Control Systems (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention discloses an accident detection method for recognizing vehicle collision through monitoring videos. The method includes the following steps that firstly, stop behaviors in a target area are determined; secondly, the surrounding historical optical flow direction and the amplitude of the stop behaviors are calculated through an inter-frame gap, whether optical flow collision points exist or not around the stop behaviors is judged, if the optical flow collision points exist, it is determined that collision accidents happen, the accidents are reported, judgment is finished, and otherwise, the third step is executed; thirdly, the inter-frame gap is changed, the surrounding historical optical flow direction and the amplitude of the stop behaviors are calculated again, whether the optical flow collision points exist or not around the stop behaviors is judged, if the optical flow collision points exist, it is determined that collision accidents happen, the accidents are reported, judgment is finished, and otherwise, the fourth step is executed; fourthly, the third step is repeated, if no optical flow collision point is found around the stop behaviors even the inter-frame gap exceeds 5, it is determined that no collision accident happens, and judgment is finished. The method has very high practical value.
Description
Technical field
The present invention relates to a kind of traffic accident detection method, specifically, relate to a kind of accident detection method of monitor video identification vehicle collision.
Background technology
Along with the sharp increase of city vehicle number, traffic pressure also becomes increasing, adopts the mode of video mode monitoring urban traffic conditions more and more general.But due to the One's name is legion of monitoring probe, rely on and manually cannot meet actual requirement to the mode that monitoring probe manages, employing high efficiency, intelligentized technological means replace traditional artificial supervision method extremely urgent.
For urban transportation, cause two of traffic paralysis large major reasons for blocking up and accident.And if accident processes not in time after occurring, directly traffic congestion can be caused again.At present, the Main Means of discovery traffic accident is following three classes:
1. receive a crime report.After accident occurs, victim is reported to traffic police's system by modes such as phones;
2. find during manual inspection;
3. road gets congestion, and last artificial verification when blocking up reason finds.
By the drawback of mode acquisition accident generation information of receiving a crime report clearly, all to lie in a comatose condition or a side goes into a coma in the situations such as the opposing party's hit-and-run both sides victim, the feedback of accident generation information may be extremely delayed.
Also larger drawback is had by manual inspection mode discovery accident.Because number of cameras is numerous, generally, people trade union key monitoring several main line camera situation, so will leave most camera and be in " blind " monitor state, cause the probability of discovery accident extremely low.
By feedback accident mode of blocking up, there is accident generation information feed back delayed feature equally.
In the face of above-mentioned situation, nowadays engendered that video intelligent analyzes the technology of traffic accident generation, it mainly realizes around following direction:
1., after accident occurs, the cross section vehicle flowrate quantity caused, speed reduce, and judge may occur traffic accident in current picture according to comparison historical traffic information.The execution mode of the program has two kinds, and one obtains flow based on coil mode, but construction trouble, road surface can be destroyed, and later maintenance expense can be higher.Another kind is by video mode statistic flow, but in urban environment, and when blocking up owing to producing, eclipse phenomena is very serious, and data on flows reliability will reduce greatly, and the accident information fed back by which is comparatively delayed.
2. fatigue detecting.The program mainly carries out signature analysis to driver's face and eyes, then judges and the generation of early warning traffic accident according to driving behavior is abnormal.The benefit of the program is to produce forewarning function to driver, but drawback is to need mandatoryly to install camera additional to motor vehicle requirement, and policy implementation can be subject to very big obstruction.In addition, and the traffic accident of not all produce be all due to driving fatigue produce, common crossing accident cause mainly concentrates on behaviors such as making a dash across the red light.
3. vehicle speed abrupt climatic change.What the program detected is the real-time Behavioral change of vehicle, analyzing, stop fast suddenly, and dwell time exceeding certain threshold value, just judges that it there occurs traffic accident when detecting moving target in picture successive video frames.The program possesses certain representativeness, and when really can detect that collision occurs, vehicle stops at the traffic accident in monitored picture, but stops behavior not in monitored picture if collide the vehicle caused, and so may produce undetected.In addition, the unexpected quick stopping behavior of vehicle also can mislead judgement.
Summary of the invention
The object of the present invention is to provide a kind of accident detection method of monitor video identification vehicle collision, solve the video analysis existed in prior art and judge the problem that traffic accident generation accuracy rate is low.
To achieve these goals, the technical solution used in the present invention is as follows:
An accident detection method for monitor video identification vehicle collision, comprises the following steps:
(1) the stopping behavior in target area is determined;
(2) adopt frame period to calculate surrounding's history light stream direction and the amplitude of stopping behavior, judge whether there is the light stream point of impingement around stopping behavior, if there is the light stream point of impingement, then determine to crash, report accident, judge to terminate; Otherwise, then step (3) is performed;
(3) change frame period, recalculate surrounding's history light stream direction and the amplitude of stopping behavior, judge whether there is the light stream point of impingement around stopping behavior, if there is the light stream point of impingement, then determine to crash, report accident, judge to terminate; Otherwise, then step (4) is performed;
(4) repeat step (3), until frame period is more than 5, still not stopping the discovery light stream point of impingement around behavior, then determining current not crashing, judging to terminate.
Further, the concrete grammar of described step (1) is as follows:
(1) carry out context update according to current picture, obtain foreground image;
(2) foreground image is added up, until there is doubtful piece of the target of stopping behavior, judge whether the doubtful block size of this target exceedes pre-set threshold value;
(3) doubtful piece of target exceedes pre-set threshold value, be then defined as stopping behavior.
Again further, the mode changing frame period in described step (3) is: increase frame period gradually.
Compared with prior art, the present invention has following beneficial effect:
(1) the present invention with the situation of change of light stream in picture for core benchmark, automatically identified by system, judge whether there is the point of impingement in picture, thus avoid artificial judgment collision accident inefficiency in prior art, feed back delayed and easily occur omit problem, substantially increase the accuracy rate that traffic accident video detects, improve the feedback efficiency of exception message, and then lay the foundation, for transport solution blockage problem decreases a great latency for improving traffic accident treatment efficiency.
(2) The present invention reduces the workload of traffic, alleviate the work load that traffic is patrolled and examined, for traffic monitoring, department saves manpower and materials, also improves operating efficiency simultaneously.
(3) principle of the invention is simple, and it is convenient to realize, and has very high practical value and wide application prospect.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of the present invention-embodiment.
Embodiment
Below in conjunction with drawings and Examples, the invention will be further described, and embodiments of the present invention include but not limited to the following example.
Embodiment
As shown in Figure 1, the accident detection method of monitor video identification vehicle collision disclosed by the invention, under same design principle, provides two kinds of detection algorithms, is described respectively below.
Algorithm one
1. determine the stopping behavior in target area.
(1) use mixed Gaussian background modeling mode generation background image according to video sequence, then obtain foreground image by front, background differential mode;
(2) foreground image is added up, until there is doubtful piece of the target of stopping behavior, judge whether the doubtful block size of this target exceedes pre-set threshold value;
(3) doubtful piece of target exceedes pre-set threshold value, be then defined as stopping behavior.
2. adopt frame period to calculate surrounding's history light stream direction and the amplitude of stopping behavior, judge whether there is the light stream point of impingement around stopping behavior, if there is the light stream point of impingement, then determine to crash, report accident, judge to terminate; Otherwise, then step 3 is performed.
Namely the so-called light stream point of impingement refers in light stream to there is the relative or crossing situation in two or more direction.If existed, then indicate that this light stream belongs to the light stream point of impingement, otherwise, then do not belong to the light stream point of impingement.
3. increase frame period, recalculate surrounding's history light stream direction and the amplitude of stopping behavior, judge whether there is the light stream point of impingement around stopping behavior, if there is the light stream point of impingement, then determine to crash, report accident, judge to terminate; Otherwise, then step 4 is performed.
4. repeat step 3, until frame period is more than 5, still not stopping the discovery light stream point of impingement around behavior, then determining current not crashing, judging to terminate.
Algorithm two
1., by the light stream situation of anchor-frame compartment of terrain detecting current picture, judge whether current picture exists the point of impingement.
Specifically, be judge whether current picture exists the point of impingement according to the change of light stream direction and amplitude in detecting current picture.
If 2. picture exists the point of impingement, then judge whether original light stream trend of point of impingement twocouese changes face to face.If original light stream trend of point of impingement twocouese all changes, then determine to crash, report accident, judge to terminate; Otherwise, then next step is performed.
Such as, the position of horizontal direction light stream should be there is, after the point of impingement occurs, become vertical direction light stream etc., occur this change, namely show to there occurs collision accident.
3. judge whether there is stopping behavior in the unit interval around the point of impingement, if existed, then determine to crash, otherwise, then determine current not crashing, judge to terminate.
Unit interval described in the present embodiment is changeable, as 1 second, 1 minute, 10 minutes etc., can adjust temporarily.
Above-mentioned two kinds of algorithms can isolated operation, uses separately, also can be combined with each other, run simultaneously, result cross-application.When there is stopping behavior, then reviewing history light stream situation, although the different situation of interval frame may be there is, judging whether that the logic of colliding is identical.When there is collision behavior, preferential judgement collision rift light stream situation of change, if collision situation has changed the traffic direction of original two targets, just need not verify stopping behavior further, otherwise need to increase stopping behavior judgement, to prove collision accident.
Above-mentioned two kinds of algorithms can be implemented separately, also can two kinds of algorithm combination implement.
Above-mentioned detection algorithm can adopt the existing monitor video of current traffic police, carries out background analysis detection by analyzer, need not destroy original pavement structure; Further, the corresponding data that detect can provide by the existing platform interface of traffic police, reduce secondary and drop into and maintenance cost, have very high practical value and application value.
Above-described embodiment is only the preferred embodiments of the present invention, not limiting the scope of the invention, as long as adopt design principle of the present invention, and the change carried out non-creativeness work on this basis and make, all should belong within protection scope of the present invention.
Claims (3)
1. an accident detection method for monitor video identification vehicle collision, is characterized in that, comprise the following steps:
(1) the stopping behavior in target area is determined;
(2) adopt frame period to calculate surrounding's history light stream direction and the amplitude of stopping behavior, judge whether there is the light stream point of impingement around stopping behavior, if there is the light stream point of impingement, then determine to crash, report accident, judge to terminate; Otherwise, then step (3) is performed;
(3) change frame period, recalculate surrounding's history light stream direction and the amplitude of stopping behavior, judge whether there is the light stream point of impingement around stopping behavior, if there is the light stream point of impingement, then determine to crash, report accident, judge to terminate; Otherwise, then step (4) is performed;
(4) repeat step (3), until frame period is more than 5, still not stopping the discovery light stream point of impingement around behavior, then determining current not crashing, judging to terminate.
2. the accident detection method of a kind of monitor video identification vehicle collision according to claim 1, it is characterized in that, the concrete grammar of described step (1) is as follows:
(1) carry out context update according to current picture, obtain foreground image;
(2) foreground image is added up, until there is doubtful piece of the target of stopping behavior, judge whether the doubtful block size of this target exceedes pre-set threshold value;
(3) doubtful piece of target exceedes pre-set threshold value, be then defined as stopping behavior.
3. the accident detection method of a kind of monitor video identification vehicle collision according to claim 2, is characterized in that, the mode changing frame period in described step (3) is: increase frame period gradually.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510614372.8A CN105208339A (en) | 2015-09-24 | 2015-09-24 | Accident detection method for recognizing vehicle collision through monitoring videos |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510614372.8A CN105208339A (en) | 2015-09-24 | 2015-09-24 | Accident detection method for recognizing vehicle collision through monitoring videos |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105208339A true CN105208339A (en) | 2015-12-30 |
Family
ID=54955738
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510614372.8A Pending CN105208339A (en) | 2015-09-24 | 2015-09-24 | Accident detection method for recognizing vehicle collision through monitoring videos |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105208339A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105206055A (en) * | 2015-09-24 | 2015-12-30 | 深圳市哈工大交通电子技术有限公司 | Accident detection method for recognizing vehicle collision through traffic monitoring video |
CN107507445A (en) * | 2017-08-17 | 2017-12-22 | 千寻位置网络有限公司 | The method for reporting traffic accident and congestion track automatically based on high accuracy positioning |
CN108921028A (en) * | 2018-06-01 | 2018-11-30 | 上海博泰悦臻电子设备制造有限公司 | The scene of a traffic accident regards acquisition method and system |
CN109934075A (en) * | 2017-12-19 | 2019-06-25 | 杭州海康威视数字技术股份有限公司 | Accident detection method, apparatus, system and electronic equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120082381A1 (en) * | 2010-09-30 | 2012-04-05 | Kabushiki Kaisha Toshiba | Video analysis method and system |
CN102496000A (en) * | 2011-11-14 | 2012-06-13 | 电子科技大学 | Urban traffic accident detection method |
CN105206055A (en) * | 2015-09-24 | 2015-12-30 | 深圳市哈工大交通电子技术有限公司 | Accident detection method for recognizing vehicle collision through traffic monitoring video |
-
2015
- 2015-09-24 CN CN201510614372.8A patent/CN105208339A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120082381A1 (en) * | 2010-09-30 | 2012-04-05 | Kabushiki Kaisha Toshiba | Video analysis method and system |
CN102496000A (en) * | 2011-11-14 | 2012-06-13 | 电子科技大学 | Urban traffic accident detection method |
CN105206055A (en) * | 2015-09-24 | 2015-12-30 | 深圳市哈工大交通电子技术有限公司 | Accident detection method for recognizing vehicle collision through traffic monitoring video |
Non-Patent Citations (2)
Title |
---|
BHASKAR, PK: "《Enhanced and Effective Parallel Optical Flow Method for Vehicle Detection and Tracking》", 《INTERNATIONAL SYMPOSIUM ON MATHEMATICAL SCIENCES AND COMPUTING RESEARCH 》 * |
张利平: "《基于光流的运动车辆检测和跟踪技术的研究》", 《车辆与动力技术》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105206055A (en) * | 2015-09-24 | 2015-12-30 | 深圳市哈工大交通电子技术有限公司 | Accident detection method for recognizing vehicle collision through traffic monitoring video |
CN105206055B (en) * | 2015-09-24 | 2018-09-21 | 深圳市哈工大交通电子技术有限公司 | A kind of accident detection method of Traffic Surveillance Video identification vehicle collision |
CN107507445A (en) * | 2017-08-17 | 2017-12-22 | 千寻位置网络有限公司 | The method for reporting traffic accident and congestion track automatically based on high accuracy positioning |
CN109934075A (en) * | 2017-12-19 | 2019-06-25 | 杭州海康威视数字技术股份有限公司 | Accident detection method, apparatus, system and electronic equipment |
CN108921028A (en) * | 2018-06-01 | 2018-11-30 | 上海博泰悦臻电子设备制造有限公司 | The scene of a traffic accident regards acquisition method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105206055A (en) | Accident detection method for recognizing vehicle collision through traffic monitoring video | |
CN101799985B (en) | Highway tunnel traffic identification method | |
KR101095528B1 (en) | An outomatic sensing system for traffic accident and method thereof | |
CN102945603B (en) | Method for detecting traffic event and electronic police device | |
CN102568206B (en) | Video monitoring-based method for detecting cars parking against regulations | |
CN103258429B (en) | Video detecting method aims at vehicles which enter into jammed intersection by force | |
CN105208339A (en) | Accident detection method for recognizing vehicle collision through monitoring videos | |
CN103985255B (en) | Radar tracking technology is utilized precisely to judge the system and method for runway exception parking | |
CN101707005B (en) | Rapidly alarming system to signal light crossing congestion by video detection and operation method thereof | |
CA2925843C (en) | System for traffic behaviour surveillance | |
CN208335464U (en) | A kind of traffic events monitoring device based on video and radar | |
CN104809887A (en) | Detecting method for vehicle converse running on expressway and automatic alarm device | |
CN103903465A (en) | Method and system for releasing reason for road congestion in real time | |
CN113139482A (en) | Method and device for detecting traffic abnormity | |
CN104282158A (en) | Intelligent control method and system for traffic lights | |
CN103150899B (en) | The monitoring method of Traffic Information, Apparatus and system | |
CN103236181A (en) | Traffic signal lamp state monitoring system and method based on machine vision | |
CN102426783B (en) | Low flow road traffic incident detection method based on vehicle tracking | |
CN114387785A (en) | Safety management and control method and system based on intelligent highway and storable medium | |
CN111754786A (en) | System for identifying traffic vehicle passing events on highway | |
CN115049993A (en) | Vehicle abnormal stop monitoring method based on deep learning | |
CN104123840A (en) | Video detection technology based traffic density qualitative judging method | |
CN105405297A (en) | Traffic accident automatic detection method based on monitoring video | |
CN114038195A (en) | Road abnormal event monitoring and detecting method based on trajectory analysis | |
CN113111824B (en) | Real-time pedestrian crossing road identification method based on video analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20151230 |