AR Marker Aided Obstacle Localization System for Assisting Visually Impaired

Abstract

As of in 2017, approximately 3.3% of the people in the world (253 million) are visually impaired. These people face mobility difficulties which impact their quality of life. We propose a system that will assist the visually impaired with their indoor environment mobility. It uses the information from preregistered AR (Augmented Reality) markers to identify specific accessible facilities, such as hallways, restrooms, staircases, and offices. An RGB-D sensor (a sensor that provides color and depth information of every pixel) captures the scene which includes the AR markers. This scene-based information is processed by a neural network to recognize and localize obstacles and accessible facilities. We present the processing algorithm for image and depth profiling. System performance in terms of obstacle localization and recognition was evaluated inside building.