CANSS: A Candidate Selection and Search Algorithm to Initialize Car Tracking

This report describes an algorithm to find cars in images taken by a forward-looking camera, mounted on a pursuing vehicle. It was developed in order to initialize a car tracker described elsewhere [2]. The algorithm proceeds in two steps, in order to find a bounding box that most probably corresponds to a car to be tracked. In a candidate selection step, potential edges for each of the 4 sides of the bounding box are selected. This is done using non-parametric density estimation followed by detection of local maxima. Then, in the search step, the most probable bounding box is selected from a set of hypotheses obtained by combining the candidate edges in all possible ways. The resulting algorithm is simple, fast, and works well in practice. This work was supported in part by USDOT under Cooperative Agreement Number DTFH61-94-X-00001 as part of the National Automated Highway System Consortium. The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing official policies or endorsements, either expressed or implied, of the United States Government or any of the sponsoring institutions.