A method to detect steps on the sidewalks for supporting visually impaired people in walking

Abstract

When visually impaired people go out, small steps and holes on the sidewalks can cause injuries to them. For example, when walking on a sidewalk, a small step is often seen at the boundary between the sidewalk and the road. People with visual disabilities often go out using a white cane, but it is predicted that using a white cane cannot help them detect a step. In such a case, there is a risk of falling even with a small step. One of the previous studies aimed at supporting visually impaired people in walking is focusing on navigation support for visually impaired people in terrain recognition using smart glasses. In this study, when detecting various objects and space, a method of detecting various objects by performing processing in real time by reducing the calculation cost is adopted. However, previous studies have not focused on detecting small steps from image data. Therefore, in order to assist visually impaired people to walk on sidewalks and roads, we aim to detect steps on the sidewalk.