Real-time traffic cone detection for autonomous vehicle
Yong Huang, Xue Jianru · 2015
Traffic signs recognition is a basic task for autonomous vehicle. Among the numerous traffic signs, traffic cone is a very important mark used to guide cars where to go. A method based on vision and radar sensors information fusion is proposed to detect traffic cone in this paper. The algorithm mainly includes two parts: finding where the obstacle is in the image and recognizing whether it is a cone. We use homography to calibrate camera and radar, from which the radar data can be mapped on the image and a small corresponding image patch can be easily cutout. Then, a method based on contour feature called chamfer matching is used to determine whether the obstacle in the image patch is a cone. The approach has been tested on our autonomous vehicle, which shows it can guarantee both effectiveness and instantaneity.