Real-time multi-platform pedestrian detection in a heavy duty driver assistance system

Reliable and fast pedestrian detection in mobile systems is an impor- tant and increasing field of research. The main goal is to save lives of pedestrians in as many situations and environments as possible. Most of the existing reliable pedestrian detection algorithms run on high performance PCs with intense CPU and GPU usage and are therefore not suited for mobile embedded systems. We present a heavy duty multi-platform mobile system running a real-time pedestrian detection. Our approach is to run each part of our software on the hardware part which is suited best for solving the problem. This distribution of the software to the several hardware parts is work in progress, but first experiments with the basic algorithm already show promising results.