Multi-lane Vehicle Detection For Traffic Intersection

With the increase of city vehicles, improving traffic congestion status has become a major goals for urban transport system development. Intelligent traffic system is a popular research direction in the future of city traffic management and traffic control. The system can be more efficient on traffic flow limit, adjustment and unplugging to improve traffic efficiency and ensure the traffic safety. Thereamong, multi-lane vehicle detection based video is the important constituent of the intelligent traffic system realization.This paper introduces some domestic and foreign driveway partition and vehicle detection technology in common use, proposes a set vehicle detection method based on the background subtraction method and the multi-lane vehicle detection method, after analyzing and comparing. The method research focuses on multi-lane dipartition and vehicle detection from the background. After the image pretreatment about the frame collected by Camera, such as gray, image enhancement and so on, frames are divided into several interested regions. In the small region, has the Hough line detection and linear features highlight. Hough line detection is one of the most mature and most widely used algorithm for line detection. Then to detect the lane lines for reliable points selection, fit out a plurality of lanes through the lane line model. Among these algorithms, the interested region doesn't only reduce the amount of computation, but also reject noise effect on lane detection, with quick response, good robust characteristic. After getting the straight lines of the lane and road boundary, to make lane division and highlight the target lane, aim to convenient vehicle single lane detection and make the traffic system more humanized and intelligent.The vehicle detection based on the background frame subtraction method, the establishment and update of background is the key to the ability to accurately detect the vehicle information. This paper introduces some common methods about the background establish and update at first, and then select the averaging method to establish the initial background, combined with the environment features of intersection using, use the Kalman filtering, introduce new factor for background update. After background subtraction, use OTSU threshold segmentation method to extract the vehicle on the single lane. This paper gives the pictures of the background establishment and learning process. The background learning rapid is very fast, and the update is also very stable, good real-time.This algorithm debugging platform is VC6.0, OpenCV1.0is embed in the implementation. The OpenCV package function code is concise, high-speed, and ideal effect.Multi-Lane vehicle detection can be used to process the data frame, at the same time, the lane identification division is more appropriate to meet three-dimensional and diversified needs in modern social transportation, improve the intelligent traffic light speed and recognition performance, reduce the cost, save energy and protect environment, response to the state of low carbon life appeal.