Research Of The Key Technology And System Implementation In Traffic Violation Video Detection

With the rapid development of road traffic,the frequency of the occurrence of traffic violations has increased,and the harm caused has also aggravated.Therefore,it is very important to effectively limit the occurrence of traffic violations.Intelligent transportation system is an important way to deal with traffic violations today which has features like wide range and high efficiency.The key technology for the video detection of traffic violation in intelligent transportation systems is focused and the system implementation is completed.The system studied in this article deals with the instant image data returned by the surveillance camera.In order to solve the problem of the program's computational efficiency brought by the large resolution of the video image and reserve the main information of the image,an extraction operation is performed for each frame of image by row and column,respectively.In order to identify the target more accurately and reduce the adverse effects of noise and other interference factors,the morphological opening operation is used in the image.The background is established using the mean method and the foreground is extracted by differential binarization.The connected domain is analyzed to identify the moving objects in the scene,and each object is represented by a rectangular class.The lighting in the monitoring scene will cause cast shadow.To solve the problem of shadow misplacing the road marking,a shadow removal algorithm based on gradient projection and wavelet transform is used to achieve the effect of removing the shadow area and segmenting the vehicle target.A shape correction was made to the target.In order to match the same moving target between different frames,the maximum overlap area threshold method is used to track the vehicle target,and the position,size,and other information of each target in the scene is recorded.And the complete travel trajectory of the target is saved.For complex monitoring scenarios,the least square method is used to fit boundary information such as monitoring area boundaries,road markings,and no-parking areas.The data is saved for reading next time by the program.After analyzing the movement trajectory of the moving target and combining the road data,the criteria for judging illegal behaviors such as riding and rolling road marking and parking in the prohibited area is defined.A motion trajectory prediction method based on pixel coordinate system that does not require actual road information and camera parameters is used.This method is based on the historical running trajectory of the moving object in the pixel coordinate system.The calculation is simple and accurate,and the prediction method is used to capture the vehicles with illegal behaviors and save image and video evidence.Digital watermarking is used to encrypted image evidence and prevented it from being tampered with.Image evidence along with video evidence are stored locally and transmitted to the server of the alarm information receiving center via the network.A complete traffic violation video detection system is designed and implemented.The system's hardware devices include dome cameras,computers and servers.The software part is based on the MFC software platform,including camera interfaces,image processing,violation identification,capturing and storing evidence,and evidence uploading modules.The camera's automatic monitoring,capturing,and uploading are realized.The system was tested in roads with high frequency of violation in real life to detect violations such as rolling markings and illegal parking.The test results were good and the system passed the product certification of the Product Quality Supervision Center of Ministry of Public Security.