High-Speed Recognition of Pedestrians out of Blind Spot with Pre-detection of Potentially Dangerous Regions

Jiacheng Zhou, Masahiro Hirano, Yuji Yamakawa

Abstract: This paper presents a novel approach for the early detection of pedestrians crossing out of blind spots. We used a monocular camera mounted at the front of a vehicle and developed an algorithm for detecting blind spots caused by obstacles blocking the view of the driver, based on the depth information obtained from a monocular depth estimation before the vehicle approaches a potentially dangerous region. Unlike other studies in which pedestrians were searched for over the entire image, our blind spot pre-detection method allows for a considerably smaller search area for pedestrian detection, which can significantly reduce the time latency of detecting unexpected pedestrians. We evaluated the proposed method on test videos with complex pedestrian and street scenes. The experimental results revealed that the performance of the method in detecting blind spots and sudden appearance of pedestrians is good, and the reaction latency for such emergency situations is significantly reduced by applying the pre-detection method. Our method can be applied in advanced driver-assistance systems and reduce accident rate caused by sudden pedestrian crossing.