Highway overhead structure detection using video image sequences

A radar-based forward collision warning system has the problem of not being able to distinguish a stopped vehicle from the false alarms caused by highway overhead structures. We describe a system that detects the presence of overhead structures based on video images from an in-vehicle camera. The system also estimates the height and distance of the structures to the host vehicle, which can be used to help reject false alarms from radar caused by overhead structures. Our system uses a unique horizontal edge projection (HEP) as the primary feature for the detection of the overhead structures, which is computationally simple and robust to noise. Image vertical motion compensation and HEP tracking algorithms are developed to establish the HEP tracks of trajectories of the overhead structures in the image sequence. A least-squares algorithm is presented for the estimation of structure height and distance to the host vehicle using the HEP tracks. Experiments using real image sequences are presented.