Fast pedestrian detection from a moving vehicle

Abstract

This paper presents a method of real-time multi-modal pedestrian detection from a moving vehicle. The system uses both intensity and thermal images captured from cameras mounted at the front of the vehicle to train cascades of classifiers, which results in a detector that is able to detect a large percentage of pedestrians with very few false positives. The system has also been tested with inputs of high-resolution intensity images along with low-resolution thermal images, showing that the addition of even a low-resolution thermal camera may return better pedestrian detection results than using only intensity information alone.