Real-Time Mobile Navigation Assistancein Unknown Indoor Environmentsfor Visually Impaired People Using A Monocular Camera

Abstract

Aplethora of solutions have been developed to cater to the inability ofvisually-impairedpeople to perceive the environment in a natural and undeterred manner. Although such techniques provide visual assistance in open environments where highly accurate sensor modalities like global positioning systems (GPS) can prove to be a reliable source of information, they fail to perform consistently in unknown indoor environments. In this work, we propose a novel approach to perform real-time mapping of unknown indoor environments through non-intrusive mobile devices using only a single monocular camera. In addition, we focus on generating high-fidelity 2D occupancy grid maps from 3D point cloud data by seamless integration with the ROS middleware to provide real-time navigation assistance by leveraging existing standard motion planning algorithms.