Pedestrian Crossing Lights and Obstacles Detections for Visually Impaired Person

Abstract

This research aims to develop a complete portable aid for visually impaired persons. It focuses on detecting the pedestrian traffic lights and obstacles in the way of a visually impaired person who crosses a pedestrian crossing. The system recognizes the pedestrian traffic lights in real-time with deep learning and provides signals to the visually impaired person to cross the pedestrian crossing in the form of audio output. It should work in both the day and night time and have good reliability. For the good reliability of the system, different sorts of images and video data have been collected, edited, and stored in a large amount of training data to get a good model.