Recognition of Obstacles from a Spherical Camera Image Based on YOLOv3

Abstract

The attention has been focused in using assistive devices at the aging era of Japan. One of the devices is electric wheelchair, which enables physical disability people to easily operate it. However, accidents are occurring frequently with increasing demand by using electric wheelchair. Therefore, the development of an autonomous driving electric wheelchair is required to reduce accidents. In this paper, we propose a recognition of obstacles of panoramic images that obtained from a spherical camera. A spherical camera is equipped in an electric wheelchair, and images are cut out from the sequential images obtained by running. For obstacles recognition, we use YOLOv3. The proposed method considers the distortion of the image caused by using the spherical camera. The improvement of the model of YOLOv3 is examined, and the validity with the actual data is verified.