A Warping Window Approach to Real-time Vision-based Pedestrian Detection in a Truck's Blind Spot Zone

We present a vision-based pedestrian tracking system targeting a very specific application: avoiding accidents in the blind spot zone of trucks. Existing blind spot safety systems do not offer a complete solution to this problem. Therefore we propose an active alarm system, which warns the truck driver if vulnerable road users occur in the blind spot zone. Our system is based solely on a vision sensor, and automatically detects vulnerable road users in the blind spot camera images. Due to the nature of this specific problem, this is a challenging task. Besides the demanding time constraint there is a need for a high accuracy, and we have to cope with the large distortion that a blind spot camera introduces. To achieve this we propose a warping windowmulti-pedestrian tracking algorithm. Our algorithm achieves real-time performance while maintaining high accuracy. To evaluate our algorithms we recorded several datasets with a real blind spot camera mounted on a real truck, consisting of realistic simulated dangerous blind spot situations.