CN115880916A - Road traffic accident detection method based on behavior model - Google Patents
Road traffic accident detection method based on behavior model Download PDFInfo
- Publication number
- CN115880916A CN115880916A CN202211490654.8A CN202211490654A CN115880916A CN 115880916 A CN115880916 A CN 115880916A CN 202211490654 A CN202211490654 A CN 202211490654A CN 115880916 A CN115880916 A CN 115880916A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- camera
- vehicles
- information
- traffic accident
- 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 discloses a road traffic accident detection method based on a behavior model, belonging to the technical field of road traffic management, and comprising the following steps: inputting the road information provided with the camera into a processor control panel of the camera, identifying the dividing line of the road surface through the camera, and acquiring the flow information of pedestrians and vehicles on the road in real time; according to the invention, the data sorting module sorts the information captured by the camera, such as the information of the illegal motor vehicles and the non-motor vehicles, and the facial information of the non-motor vehicle riding personnel and the illegal pedestrians, and then transmits the sorted information to the electric police terminal server through the switch, the video is identified through the cloud processor, and the illegal information is fed back to the electric police terminal server; the data are transmitted to a display of the illegal intersection to be published, and a warning effect is achieved on subsequent vehicles, so that the probability of traffic accidents caused by illegal traffic is reduced.
Description
Technical Field
The invention relates to the technical field of road traffic management, in particular to a road traffic accident detection method based on a behavior model.
Background
With the rapid development of urban roads and the gradual improvement of the automobile holding capacity of residents, people enjoy the trouble of various traffic accidents while enjoying the comfortable experience brought by automobiles. Traffic accidents threaten the life safety of the parties and influence the normal traffic of other vehicles on the road. The responsibility determination of the vehicle after the accident is also a big problem. In a traffic supervision and management system, casualties and property losses of related personnel can be reduced by accurately identifying traffic accidents occurring on a road, so that the detection of the traffic accidents on the road has great significance.
At present, a traffic accident detection method based on video mainly utilizes an image pyramid of a video frame to construct an optical flow field. And calculating information such as velocity flow, acceleration flow and the like from the optical flow field, and judging whether a traffic accident occurs or not by setting a threshold value. The learner also detects the vehicle by using the Faster-RCNN, judges whether the vehicle is stationary in consecutive frames, and roughly estimates whether an abnormal accident occurs to the vehicle according to the vehicle stop time. In addition, there is a method of directly calculating the vehicle speed and determining whether a traffic accident occurs according to whether the vehicle speed is abnormal.
In the existing method, the speed of each vehicle cannot be accurately calculated when a large number of vehicles exist in the method for calculating the optical flow field by directly using the image pyramid, and the speed flow only can be roughly estimated for the vehicles in the video scene. Meanwhile, the method for judging whether a traffic accident occurs according to the manually set threshold value seriously depends on subjective thoughts and has no flexibility in practical application.
The method of determining whether a vehicle stops and calculating the stop time to determine whether a traffic accident occurs also requires setting the maximum stop time of the vehicle. In the case of traffic jam or waiting for a turn on the actual road, the method may be misreported, and therefore, the identification of the traffic accident is not accurate enough.
For the method for counting the vehicle speed and judging whether a traffic accident occurs or not according to the vehicle speed, the accuracy is still low, and the main reason is that the driving condition of the vehicle in a period of time cannot be fully reflected only by using single information of instantaneous vehicle speed;
the prior patent grant publication numbers are: CN113378803B discloses a method for detecting traffic accidents, which specifically includes inputting video frames, identifying vehicles in the video, and calculating vehicle speed, vehicle acceleration and vehicle head heading angle; secondly, converting time domain signals of the vehicle speed, the vehicle acceleration and the heading angle of the vehicle head into time frequency signals; secondly, inputting the time-frequency diagram into a Resnet50 network for training; and finally, extracting characteristics, and comparing the characteristics of the time-frequency graph of the vehicle speed, the vehicle acceleration and the vehicle head orientation angle with the characteristics of the time-frequency graph of the vehicle speed, the vehicle acceleration and the vehicle head orientation angle under normal conditions to judge the occurrence of traffic accidents. The technical problems that the road traffic accident detection effect is unstable and the recognition effect is not accurate in the prior art are solved. The method and the device realize accurate judgment of the road traffic accidents, and have stronger applicability and generalization capability.
However, the above technical solutions are to predict the traffic accident early warning simulation between the opposite vehicles, and it is impossible to predict the vehicles in the same direction, and there are lane changing without turning on the turn signal or continuous lane changing between the vehicles in the same direction, and the vehicles are not turned on or off according to regulations, for example, the lights are not turned on at night and the eye lights are illegally used, so that the application range is narrowed, and the use efficiency is affected.
Therefore, the invention discloses a road traffic accident detection method based on a behavior model.
Disclosure of Invention
The present invention has been made in view of the above and/or other problems occurring in the conventional behavior model-based road traffic accident detection method.
Therefore, an object of the present invention is to provide a method for detecting a road traffic accident based on a behavior model, which can solve the problems that the application range is narrowed and the use efficiency is affected, because the traffic accident early warning simulation between opposite vehicles is predicted in the prior art, the prediction cannot be performed between the vehicles in the same direction, the lane change is not performed, or the lane change is performed continuously, the vehicles are not turned on or off according to the regulations, for example, the lights are not turned on at night, and the eye lights are used illegally.
To solve the above technical problem, according to an aspect of the present invention, the present invention provides the following technical solutions:
a method of road traffic accident detection based on a behaviour model, comprising: the method comprises the following steps:
s1: inputting the road information provided with the camera into a processor control panel of the camera, identifying the dividing line of the road surface, the road surface mark and the traffic light signal through the camera, and acquiring the track information of pedestrians and vehicles on the road in real time;
s2: the video information acquired by the camera is sorted by the data sorting module and then transmitted to the electric alarm terminal server through the switch, the data is transmitted to the cloud processor through the electric alarm terminal server, the video is identified through the cloud processor, and the illegal information is fed back to the electric alarm terminal server;
s3: the violation information is issued through the electric warning terminal server, the data are transmitted to the displayer of the violation intersection to be published, and the follow-up vehicle is warned, so that the probability of traffic accidents caused by violation is reduced.
As a preferable scheme of the road traffic accident detection method based on the behavior model according to the present invention, wherein: the method comprises the steps of acquiring traffic lights and information of the ramp of the intersection while receiving screen data transmitted back by the camera through the switch, matching the information with the transmitted screen data, and providing time and indication data basis of the traffic lights of the intersection for subsequent violation judgment.
As a preferable scheme of the behavior model-based road traffic accident detection method of the present invention, wherein: the camera captures the lane-changing vehicle, identifies and captures whether the lane-changing vehicle turns on a turn light, and simultaneously captures and judges whether the lane-changing vehicle continuously changes.
As a preferable scheme of the behavior model-based road traffic accident detection method of the present invention, wherein: the camera captures the plugged vehicles, and vehicles which are not lighted at night and illegally use high beams are captured.
As a preferable scheme of the road traffic accident detection method based on the behavior model according to the present invention, wherein: the camera captures red light running of the non-motor vehicle, backward running of the non-motor vehicle and helmet wearing-free of the non-motor vehicle, and captures face data of a rider driving the non-motor vehicle.
As a preferable scheme of the road traffic accident detection method based on the behavior model according to the present invention, wherein: the camera collects the face data of the pedestrians running the red light, and the motor vehicles catch the vehicles of the pedestrians without courtesy.
As a preferable scheme of the behavior model-based road traffic accident detection method of the present invention, wherein: and collecting the captured vehicle information, the time tag and the moving speed of the violation vehicle.
As a preferable scheme of the road traffic accident detection method based on the behavior model according to the present invention, wherein: illegal video information captured by the camera is transmitted to an electric police middle-end server through a switch electrically connected with the camera.
As a preferable scheme of the road traffic accident detection method based on the behavior model according to the present invention, wherein: inputting a video frame to the camera, identifying vehicles in the video, and calculating vehicle speed, vehicle acceleration and vehicle head direction angle; the method specifically comprises the following steps: identifying vehicles in the video frame and outputting a vehicle detection frame; tracking the position of the vehicle in the continuous frames to obtain a detection frame of the movement of the same vehicle in the continuous frames; calculating the pixel distance of the vehicle movement between two frames by taking the detection frame as the vehicle center; calculating a vehicle speed and a vehicle acceleration; and judging the direction angle of the vehicle head according to the angle of the detection frame.
As a preferable scheme of the road traffic accident detection method based on the behavior model according to the present invention, wherein: converting time domain signals of vehicle speed, vehicle acceleration and vehicle head direction angle into a time frequency graph; inputting the time-frequency graph into a Resnet50 network for training; preparing training data, selecting Q roads, then selecting video data of a plurality of time periods at different angles on the roads, and dividing the video data into video data of one minute; processing each segmented video to obtain a time-frequency image data set; dividing a time-frequency graph data set into a training set and a verification set; calculating the distance lambda of the characteristic vector, wherein the distance lambda of the characteristic vector is the maximum distance of the characteristic vector of the verification set and the characteristic vector of the training set; and S4, comparing the time-frequency graph characteristics of the vehicle speed, the vehicle acceleration and the vehicle head direction angle with the time-frequency graph characteristics of the vehicle speed, the vehicle acceleration and the vehicle head direction angle under the normal condition, and judging the occurrence of the traffic accident.
Compared with the prior art:
the data sorting module sorts the information of the illegal motor vehicles and the non-motor vehicles acquired by the camera and the information captured by the facial information of the riding personnel of the non-motor vehicles and the illegal pedestrians, and then transmits the sorted information to the electric alarm terminal server through the switch, the electric alarm terminal server transmits the data to the cloud processor, the cloud processor identifies videos, and the illegal information is fed back to the electric alarm terminal server; the violation information is issued through the alarm terminal server, the data is transmitted to a display of the violation intersection to be issued, and a warning effect is exerted on subsequent vehicles, so that the probability of traffic accidents caused by violation is reduced.
Drawings
FIG. 1 is a circuit flow diagram of the present invention;
FIG. 2 is a flow chart of the method of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The invention provides a road traffic accident detection method based on a behavior model, which has the advantages of enlarging the application range and improving the detection efficiency, and please refer to fig. 1-2, and comprises the following steps:
s1: inputting the road information provided with the camera into a processor control panel of the camera, identifying the dividing line of the road surface, the road surface mark and the traffic light signal through the camera, and acquiring the track information of pedestrians and vehicles on the road in real time;
s2: the video information acquired by the camera is sorted by the data sorting module and then transmitted to the electric alarm terminal server through the switch, the data is transmitted to the cloud processor through the electric alarm terminal server, the video is identified through the cloud processor, and violation information is fed back to the electric alarm terminal server;
s3: the violation information is issued through the alarm terminal server, the data is transmitted to a display of the violation intersection to be issued, and a warning effect is exerted on subsequent vehicles, so that the probability of traffic accidents caused by violation is reduced.
The method comprises the steps of acquiring traffic lights and information of the ramp of the intersection while receiving screen data transmitted back by the camera through the switch, matching the information with the transmitted screen data, and providing time and indication data basis of the traffic lights of the intersection for subsequent violation judgment.
The camera captures the lane-changing vehicle, and simultaneously identifies and captures whether the lane-changing vehicle turns on a turn light, and simultaneously captures and judges whether the lane-changing vehicle continuously changes.
The camera captures the plugged vehicle, and captures the vehicle which is not lighted at night and illegally uses the high beam.
The camera captures red light running of the non-motor vehicle, backward running of the non-motor vehicle and helmet wearing-free of the non-motor vehicle, and captures face data of a rider driving the non-motor vehicle.
The camera collects the face data of the pedestrians running the red light, and the motor vehicles catch the vehicles of the pedestrians without courtesy.
And collecting the captured vehicle information, the time tag and the moving speed of the violation vehicle.
Illegal video information captured by the camera is transmitted to an electric police middle-end server through a switch electrically connected with the camera.
Inputting video frames to the camera, identifying vehicles in the video, and calculating vehicle speed, vehicle acceleration and vehicle head direction angles; the method specifically comprises the following steps: identifying vehicles in the video frame and outputting a vehicle detection frame; tracking the position of the vehicle in the continuous frames to obtain a detection frame of the movement of the same vehicle in the continuous frames; calculating the pixel distance of the vehicle movement between two frames by taking the detection frame as the vehicle center; calculating a vehicle speed and a vehicle acceleration; and judging the direction angle of the vehicle head according to the angle of the detection frame.
Converting time domain signals of vehicle speed, vehicle acceleration and vehicle head direction angle into a time frequency graph; inputting the time-frequency graph into a Resnet50 network for training; preparing training data, selecting Q roads, then selecting video data of a plurality of time periods at different angles on the roads, and dividing the video data into video data of one minute; processing each segmented video to obtain a time-frequency image data set; dividing a time-frequency graph data set into a training set and a verification set; calculating the distance lambda of the characteristic vector, wherein the distance lambda of the characteristic vector is the maximum distance of the characteristic vector of the verification set and the characteristic vector of the training set; and S4, comparing the time-frequency graph characteristics of the vehicle speed, the vehicle acceleration and the vehicle head direction angle with the time-frequency graph characteristics of the vehicle speed, the vehicle acceleration and the vehicle head direction angle under the normal condition, and judging the occurrence of the traffic accident.
While the invention has been described above with reference to an embodiment, various modifications may be made and equivalents may be substituted for elements thereof without departing from the scope of the invention. In particular, the various features of the disclosed embodiments of this invention can be used in any combination as long as there is no structural conflict, and the combination is not exhaustively described in this specification merely for the sake of brevity and resource savings. Therefore, it is intended that the invention not be limited to the particular embodiments disclosed, but that the invention will include all embodiments falling within the scope of the appended claims.
Claims (10)
1. A road traffic accident detection method based on a behavior model is characterized in that: the method comprises the following steps:
s1: inputting the road information provided with the camera into a processor control panel of the camera, identifying the dividing line of the road surface, the road surface mark and the traffic light signal through the camera, and acquiring the track information of pedestrians and vehicles on the road in real time;
s2: the video information acquired by the camera is sorted by the data sorting module and then transmitted to the electric alarm terminal server through the switch, the data is transmitted to the cloud processor through the electric alarm terminal server, the video is identified through the cloud processor, and violation information is fed back to the electric alarm terminal server;
s3: the violation information is issued through the electric warning terminal server, the data are transmitted to the displayer of the violation intersection to be published, and the follow-up vehicle is warned, so that the probability of traffic accidents caused by violation is reduced.
2. The behavior model-based road traffic accident detection method according to claim 1, wherein the traffic lights and the information of the turn-to-turn roads at the intersections are acquired while screen data returned by the cameras through the switches are received, and the screen data are matched with the returned screen data, so that time and indication data basis of the traffic lights at the intersections are provided for subsequent violation judgment.
3. The behavior model-based road traffic accident detection method according to claim 2, wherein the camera captures lane-changed vehicles, and simultaneously recognizes and captures whether the lane-changed vehicles turn on a turn signal lamp, and simultaneously performs capture judgment on whether the lane-changed vehicles continuously change.
4. The behavior model-based road traffic accident detection method according to claim 3, wherein the camera captures a jammed vehicle, and the camera captures a vehicle which is not lighted at night and illegally uses a high beam.
5. The behavior model-based road traffic accident detection method according to claim 4, wherein the cameras capture the red light violation of the non-motor vehicles, the backward driving of the non-motor vehicles and the helmet-free driving of the non-motor vehicles, and capture the face data of the riders driving the non-motor vehicles.
6. The behavior model-based road traffic accident detection method according to claim 5, wherein the camera is used for collecting face data of pedestrians running red light, and vehicles of motor vehicles which do not give way to pedestrians are captured.
7. The behavior model-based road traffic accident detection method according to claim 6, characterized in that the captured vehicle information, time labels and moving speed of the offending vehicles are collected.
8. The method for detecting the road traffic accident based on the behavioral model according to any one of claims 3 to 7, wherein the violation video information captured by the camera is transmitted to the electric police middle-end server through a switch electrically connected with the camera.
9. The behavior model-based road traffic accident detection method according to claim 8, wherein video frames are input to the camera, vehicles in the video are identified, and vehicle speed, vehicle acceleration and heading angle are calculated; the method specifically comprises the following steps: identifying vehicles in the video frame and outputting a vehicle detection frame; tracking the position of the vehicle in the continuous frames to obtain a detection frame of the movement of the same vehicle in the continuous frames; calculating the pixel distance of the vehicle movement between two frames by taking the detection frame as the vehicle center; calculating a vehicle speed and a vehicle acceleration; and judging the direction angle of the vehicle head according to the angle of the detection frame.
10. The behavior model-based road traffic accident detection method according to claim 9, characterized in that time domain signals of vehicle speed, vehicle acceleration and heading direction angle are converted into time-frequency graphs; inputting the time-frequency graph into a Resnet50 network for training; preparing training data, selecting Q roads, then selecting video data of a plurality of time periods at different angles on the roads, and dividing the video data into video data of one minute; processing each segmented video to obtain a time-frequency image data set; dividing a time-frequency graph data set into a training set and a verification set; calculating the distance lambda of the characteristic vector, wherein the distance lambda of the characteristic vector is the maximum distance of the characteristic vector of the verification set and the training set; and S4, comparing the time-frequency graph characteristics of the vehicle speed, the vehicle acceleration and the vehicle head direction angle with the time-frequency graph characteristics of the vehicle speed, the vehicle acceleration and the vehicle head direction angle under the normal condition, and judging the occurrence of the traffic accident.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211490654.8A CN115880916A (en) | 2022-11-25 | 2022-11-25 | Road traffic accident detection method based on behavior model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211490654.8A CN115880916A (en) | 2022-11-25 | 2022-11-25 | Road traffic accident detection method based on behavior model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115880916A true CN115880916A (en) | 2023-03-31 |
Family
ID=85763981
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211490654.8A Pending CN115880916A (en) | 2022-11-25 | 2022-11-25 | Road traffic accident detection method based on behavior model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115880916A (en) |
-
2022
- 2022-11-25 CN CN202211490654.8A patent/CN115880916A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11113961B2 (en) | Driver behavior monitoring | |
| US6442474B1 (en) | Vision-based method and apparatus for monitoring vehicular traffic events | |
| CN112700470B (en) | Target detection and track extraction method based on traffic video stream | |
| US11836985B2 (en) | Identifying suspicious entities using autonomous vehicles | |
| WO2018058958A1 (en) | Road vehicle traffic alarm system and method therefor | |
| EP3403219A1 (en) | Driver behavior monitoring | |
| CN109191829B (en) | road safety monitoring method and system, and computer readable storage medium | |
| CN101739809A (en) | Automatic alarm and monitoring system for pedestrian running red light | |
| CN201307337Y (en) | Automatic alarming and monitoring device for traffic-lights nonobservance of pedestrian | |
| CN111325978A (en) | Whole-process monitoring and warning system and method for abnormal behaviors of vehicles on expressway | |
| CN108932849B (en) | Method and device for recording low-speed running illegal behaviors of multiple motor vehicles | |
| CN113012436B (en) | Road monitoring method and device and electronic equipment | |
| CN113055649A (en) | Tunnel intelligent video monitoring method and device, intelligent terminal and storage medium | |
| US20240046787A1 (en) | Method And System For Traffic Clearance At Signalized Intersections Based On Lidar And Trajectory Prediction | |
| CN101349562A (en) | Method and apparatus for alarming vehicle running bias direction | |
| CN113033275B (en) | Vehicle lane-changing non-turn signal lamp analysis system based on deep learning | |
| CN112489383A (en) | Early warning system and method for preventing red light running accident based on machine vision | |
| CN112528759A (en) | Traffic violation behavior detection method based on computer vision | |
| CN111899514A (en) | Artificial intelligence's detection system that blocks up | |
| CN113870551A (en) | Roadside monitoring system capable of identifying dangerous and non-dangerous driving behaviors | |
| CN112907979A (en) | Motor vehicle illegal driving track tracking system and method based on multiple cameras | |
| CN111291722A (en) | Vehicle weight recognition system based on V2I technology | |
| CN115662154A (en) | Intelligent signal lamp system based on video monitoring | |
| CN115880916A (en) | Road traffic accident detection method based on behavior model | |
| CN113177443A (en) | Method for intelligently identifying road traffic violation based on image vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |