Road-crossing assistance by traffic flow analysis

Abstract

We present a system to alert visually impaired pedestrians of vehicles approaching a road-crossing without traffic control. The system is computationally efficient, requires low-cost hardware, and can be mounted on existing street infrastructure, such as sign or lighting poles. The incoming video stream, showing the approaching traffic, is transformed to a one-dimensional signal, that is forwarded to a decision module. Preliminary experimental results indicate promising probabilityof- detection and false alarm rates, while providing sufficiently early warning to the pedestrian. The planned target hardware is a solar-charged low cost Android device.