Exploring Early Solutions for Automatically Identifying Inaccessible Sidewalks in the Physical World using Google Street View

Poorly maintained sidewalks, missing curb ramps, and other obstacles pose considerable accessibility challenges. Although pedestrianand bicycle-oriented maps and associated routing algorithms continue to improve, there has been a lack of work focusing on accessibility. There is currently no way for a user to determine accessible areas of a city prior to travel. In this paper, we explore the use of computer vision techniques (a linear SVM) to detect sidewalk accessibility problems in Google Street View imagery. This is early work in a large on-going project. Here, we focus on automatically identifying one key accessibility barrier: the presence/absence of curb ramps. Our preliminary results point to the potential of using automatic, highly-scalable approaches to extract information about physical accessibility from online map imagery.