Infrared Image Pedestrian Detection via YOLO-V3

Abstract

Thermal imaging is often used for pedestrian detection at night to make up for the inability of traditional cameras to be used in harsh lighting conditions. However, the pedestrian detection technology based on thermal imaging is difficult to overcome the interference of high-heat objects around the human body, and it is easy to cause misidentification. This paper proposes to detect pedestrians in infrared images at night through the YOLO-V3 detection framework. Experimental results show that the proposed method can not only accurately locate pedestrians at night, but also maintain high detection accuracy when pedestrians overlap.

Topics & Concepts