The Research On Video-Based Vehicle Detection Technique For Traffic Crossing

With the rapid development of society and economy, people's living standard has gotten great improvement and the number of vehicles has quickly increased. A large amount of manpower and material resources have been devoted to the research on Intelligent Transportation Systems (ITS) in most countries in the world. Vehicle detection is the foundational part of ITS and provides basic data for the system. In so many methods on vehicle detection, video-based detection is outstanding and has several apparent advantages such as easily intervened and lower costs. It is one of the most promising new technologies for large-scale data collection and implementation of vehicle guidance. Moreover, it is also a hotspot in research field of computer vision and image processing. The video vehicle detection and tracking system is used to get traffic information, such as vehicle flow, vehicle length, vehicle velocity and the road utilization without destroying the roads. It particularly emphasizes on the management of roads such as the traffic management, road design layout etc.This paper is to design video-based vehicle detection system for traffic crossing, which can be applied to ITS and improve the management of traffic. In this paper, we can describe researches as follows:(1) Vehicle Detection Based on Video Virtual-Line Analysis: This paper improves background subtraction method and presents a new method of traffic flow measurement based on video virtual-line. This new method is adopted to simultaneously count vehicles of several tracks in the optical window without being influenced by the position of the traveling vehicle no matter whether the vehicle is on the regular track or not, what model and how big the vehicle is. In the case of mistaken and omit detection, this method has improved its efficiency by many statistics ratification measures such as preparative estimation, rectification and relative modification.(2) Muti-Kalman Filtering Algorithm for Automatic Tracking of Multi-Vehicle: An improved algorithm of Multi-Kalman filtering for Multi-vehicle tracking is proposed. The method is based on motion objects segmentation by twice-frame-difference mask images. Each kalman filter is created based on each motion object. After the region combination in the prediction estimation area according with the real size of cars in video sequence, the moving car can be tracked by the biggest motion object mask.(3) Vehicle Moving Direction Detection Algorithm Based on Results of Tracking: A new detecting method of vehicle moving direction for complicated traffic crossing was presented. To begin with, the paper discusses the improvement of frame-difference method and builds a later detecting region for video image sequence in which the image is binary. And then, a method based on morphology is used to eliminate the noise and the regional growth method is used to find the target vehicle. Finally, the paper makes use of MAD guidelines and an improved three-step search algorithm for matching to complete the detection of vehicle moving direction.According to the above-mentioned techniques, a video-based vehicle detection system for traffic crossing is developed with the function for vehicle detection and vehicle moving direction detection. The traffic flow parameters of each direction such as vehicle flow and vehicle speed can be obtained using one camera without traffic signals by the system, and the approach gets good effect.