Real-time Vehicle Detection and Tracking Algorithm for Forward Vehicle Collision Warning

초록

A Forward vehicle detection and lane detection algorithms are essential components of many advanced driver assistance systems. Conventional vehicle detection algorithms exhibit certain problems, including high computational complexity and low detection accuracy. This paper proposes a monocular vision-based vehicle detection and tracking algorithm using ego-lane information to improve computational efficiency and robustness. Firstly, vehicle candidate regions were determined to define the region of interest (ROI), which reduces the computation time. Secondly, the detected ROI was classified by means of the adaptive boosting cascade classifier, which is based on Haar-like features, in order to detect the vehicles' rear view. Thirdly, edge and rear-light histograms obtained from previously detected vehicle locations were employed to predict the area in which the vehicle is located. Experiments were conducted to evaluate the proposed algorithm under various weather and illumination conditions using iROADS [34], one of the most popular datasets. The results show that the proposed algorithm performed well in real time (computation time of 15 ms) and showed high reliability in various road conditions.

키워드