Abstract
We are aiming to build a system which provides a route with minimum risk in order to reduce accidents of elderly drivers. For this purpose， it is necessary to estimate the risk level of unsignalized intersections where accidents are likely to occur. Therefore, for estimating the risk level， we propose a method to estimate the available viewing angle at intersections based on semantic segment distribution generated from a street view image. The performance of estimation is evaluated through experiments using street view images. As a result, we achieved an accuracy of more than 80% from a single image.
