Detecting walkable plane areas by using RGB-D camera and accelerometer for visually impaired people

Abstract

When visually impaired person has to walk out, they have to use white canes, but the range that can be scanned by a white cane is not long enough to walk safely. We propose to detect walkable plane areas on road surface by using the RGB-D camera and the accelerometer in the tablet terminal that is attached to the RGB-D camera. Our approach can detect plane areas in longer distance than a white cane. It is achieved by using height information from the ground and normal vectors of the surface calculated from a depth image obtained by the RGB-D camera in real time.