Automatic Detection of Pedestrian Signals Using a Cascade Classifier

Abstract

This paper proposes a real-time detection system using Raspberry Pi and a cascade classifier to guide pedestrian signals for the visually impaired. It is desirable for the visually impaired to have a device that notifies them of the signal status at hand while they are on the move, which is easy to carry. For this purpose, it is hoped to connect a web camera to a Raspberry Pi, a small computer with excellent portability, to recognize the status of pedestrian signals and communicate it to the visually impaired. We use a cascade classifier to detect a blue signal as a preliminary investigation. Experimental results demonstrate that our system can detect the signal with precision rate of 0.909 and recall rate of 0.845 in image-based evaluation. Also, it can detect the simulated signal data with the Raspberry Pi is easy under daylight conditions, especially for LED-based signals.