An RGBD data based vehicle detection algorithm for vehicle following systems

Autonomous vehicle following system is an important research issue in the ITS(Intelligent Transportation System). After reviewing some currently used environment perception sensors, this paper employs Kinect sensor as the main device in detecting the angle and distance of the leader vehicle in relation to the following vehicle. This paper also propose a vision-based algorithm to process RGB-D data. For the color image, we use template matching and Camshift algorithm to detect and track our desired target, the result of which is to get an approximately location of the target in the image and a search window of a certain scale, both of which are used in the disposition of the depth image. We use K-means clustering to distinguish the leader vehicle and the background so that we can determine the pose information of our target. Offline simulations and online experiments have been performed to evaluate the effectiveness of the algorithm. Both of them have shown the feasibility in the vehicle following system.