Extending the detection range of vision-based driver assistance systems application to Pedestrian Protection System

Pedestrian Protection System (PPS) has become an active research area aimed to protect pedestrians by assisting inattentive drivers. Despite the enormous number of research works, most current systems are designed to recognize near-scale pedestrians. However, they perform poorly in mid-scale and fail in far-scale. In this paper, a new hardware-based architecture that detects pedestrians in near-, mid- and far-scales is introduced. Two regions are defined in this architecture: the near-to-mid area located in front of vehicles, called Region of High Risk (RHR); and the mid-to-far area, called Region of Low Risk (RLR). To implement this system, we use two identical cameras, each of which is equipped with a variable focal length lens. Moreover, we develop a mathematical model for our framework. Finally, we conduct a set of experiments in open park areas and in Ottawa roads. The results show that the system is able to accurately detect pedestrians that are located up to 130 meters away from the vehicle.