Vehicle recognition based on radar and vision sensor fusion for automatic emergency braking

In this paper, a vehicle recognition algorithm based on radar and vision sensors is proposed with the application to automatic emergency braking. While the commercial radar detects both vehicles and road infrastructure including guardrail and tunnel, in general it does not distinguish between a vehicle and a non-vehicle object. Furthermore, it is well known that while it provides relatively coarse accuracy in the lateral (or azimuth) direction although the accuracy of the radar is high in longitudinal (or radial) direction. These characteristics of radar may cause false detection of a primary vehicle, i.e. the closest preceding vehicle in the same lane, thus resulting in false activation of automatic emergency braking. To improve the false detection, a vehicle recognition method based on shape and motion attributes is suggested. The motion attribute is designed to determine whether the object is either stationary or dynamic and the shape attribute aims to identify whether the objective is a vehicle or not by sensor fusion. Finally, the performance of the proposed vehicle recognition algorithm is validated via the field test data.