Forward Collision Warning system based on vehicle detection and tracking

Recently, researches about Forward Collision Warning (FCW) technology has become one of the main parts of the development of Advanced Driver Assistance Systems (ADAS) to increase the road safety and develop autonomous vehicles. This paper presents a research about FCW based on monocular vision, which consists of two processes: vehicle detection and vehicle tracking. In the vehicle detection process, the vertical and horizontal edges and the shadows under the vehicles are combined at first to roughly generate hypothesis of all the vehicles' positions, and then HOG transform and an improved Adaboost classifier are applied to these positions to eliminate non-vehicles. In the vehicle tracking process, frames are handled by an innovative feature point matching method based on Harris detector and normalized cross correlation, and the distance is measured by checking the bottom line of newly tracked vehicles' bounding box. The presented method is verified by experiments, and is proved to have good performance.