Navigation system for visually impaired people based on RGB-D camera and ultrasonic sensor

Abstract

This paper presents a smart guiding system and wearable device to assist the visually impaired people in avoiding obstacles and navigate safely through indoor environments. The proposed navigation system includes a pair of glasses with a RGB-D camera and an ultrasonic sensor in the front, Raspberry pi 3 B+ board, and earphone. An obstacle avoiding algorithm based on integration of a computer vision and ETA techniques is adopted to provide an accurate and low cost solution. The measuring distance approach depends on an ultrasonic sensor and the raw depth map that is acquired from depth sensor to output several suitable moving directions. The integration of different data input helps to detect the small and transparent obstacles and increases the accuracy of the navigation system output. This work has been tested and evaluated in different real-time scenarios. The experimental results indicate accurate guiding instructions and effective performance of obstacle detected and avoiding algorithm. The guiding instruction has been sent to the user as audio message through earphone.