Lane Change Detection In Expressway Tunnel Based On Vehicle Light Tracking

The illegal lane change of vehicles in the tunnel is more likely to cause traffic accidents than that in the open air,and the consequences caused by the accident are more serious.Therefore,it is very important to detect and track the vehicles in real time and deal with the illegal lane change behavior in time.The traditional manual snapshot method requires a lot of manpower,material and financial resources,and the automatic illegal lane change detection based on video can greatly reduce the investment in all aspects,and has a good application prospect.In the tunnel scene,due to the influences of image quality,changes in tunnel lighting and vehicle lights,the commonly used vehicle detection methods are not ideal.As the most significant feature of the vehicle in the tunnel,it is easier to extract the vehicle light than the vehicle target.Therefore,it is of great significance to design and implement a lane change detection method of expressway tunnel based on vehicle light tracking for illegal lane change detection in tunnel scenes.For the specific scenario of expressway tunnel,after analyzing the problems and difficulties of lane change detection of expressway tunnel based on vehicle light tracking,this paper focuses on the extraction of vehicle lights in the scene of expressway tunnel,the tracking of vehicle lights in the scene of tunnel and the method of lane change discrimination,and finally forms a set of lane change detection algorithms of expressway tunnel based on the tracking of vehicle lights.The main research contents include:(1)In the aspect of vehicle light extraction,firstly,the methods of region of interest extraction and image enhancement were used to remove the noise interference in the video image of the tunnel,and the contrast between the vehicle light area and the background area was enhanced at the same time.Then,aiming at the problem of poor segmentation effect of traditional vehicle light segmentation method in expressway tunnel scene,after analyzing the characteristics of high brightness in the vehicle lights area,the vehicle lights are segmented based on gradient information.This method effectively improves the accuracy of vehicle lights extraction.Finally,aiming at the problem that the dual tail lights and high brake lights in the actual monitoring image of expressway tunnel will affect the real-time performance of the follow-up vehicle lights tracking,tail light matching and high brake light matching will be performed.(2)In the aspect of vehicle light tracking,aiming at the problems of poor light and low camera Angle of highway tunnel which lead to poor tracking effect and easy to lose tracking target,a tracking method based on feature matching and Kalman filtering is presented,which can track the target accurately and in real time.Aiming at the problem that the tunnel scene is prone to vehicle occlusion,a tracking method combining the relative position of multiple vehicles and Kalman prediction is presented.Experimental results show that this method improves the tracking accuracy of vehicle lights in tunnel scene.Finally,based on the above research results,a set of lane change detection methods for expressway tunnel based on vehicle light tracking are formed.The design and implementation were carried out under the VS2015 and OpenCV2.4.9 platforms,and the video data of the Yu-Wu expressway tunnel collected on the spot was used for experimental verification.The experimental results show that the proposed method can accurately extract the vehicle light targets and improve the accuracy of lane change detection under the premise of guaranteeing the real-time performance of the algorithm.