Fusion of infrared vision and radar for estimating the lateral dynamics of obstacles

Automotive forward collision warning systems are based on range finders to detect the obstacles ahead and warn or intervene when a dangerous situation occur. However, the radar information by itself is not adequate to predict the future path of vehicles in collision avoidance systems due to the poor estimation of their lateral attribute. In order to face this problem, this paper proposes the utilization of a new Kalman based filter, whose measurement space includes data from a radar and a vision system. Given the superiority of vision systems in estimating azimuth and lateral velocity, the filter proves to be robust in vehicle maneuvers and curves. Results from simulated and real data are presented, providing comparative results with stand alone tracking systems and the cross-covariance technique in multisensor architectures.