Detection and tracking of pedestrians in infrared images

Abstract

This article presents a pedestrian detector by means of a far infrared thermal camera, as part of an integrated driver assistance system. The final goal of the system is to warn the driver of pedestrians ahead of the vehicle. Detection is achieved by the extraction and evaluation of the main feature of pedestrians in far infrared images: the distribution of heat of the human body. Those objects with an appropriate size and temperature are correlated with probabilistic models that represent the average temperature of the different parts of the human body. Each pedestrian is tracked independently by an Unscented Kalman Filter.