Development of Pedestrian Detection System Using Far-Infrared Ray Camera

Driver assistance systems are being developed by automakers as mean to reduce traffic accidents. For examples, night vision systems that warn drivers of pedestrians during nighttime driving and pre-crash safety systems for collision mitigation have been commercialized in recent years, indicating that pedestrian detection will become an important function for driver assistance systems in the near future. Use of far-infrared ray (FIR) cameras, which offer thermographic images, is expected to improve the performance of nighttime pedestrian detection. The authors had developed FIR cameras using low cost ZnS lenses, and also developed an algorithm for detecting pedestrians during nighttime. In this paper, the authors propose a new algorithm which aims to detect pedestrians not only in the nighttime but also in the daytime and in bad weather. The paper also describes a comparison of pedestrian detection sensors, an overview of the developed FIR camera, the details of the new pedestrian detection algorithm, and the results of the tests carried on urban roads.