Smart Guide System for Blind People by Means of Stereoscopic Vision

The main goal of this work is to carry out a system that complements the tasks of detecting the white cane, locating objects or obstacles that are hollow, exceeding their detection and aerial range, through a stereoscopy system implemented in an embedded breadboard Raspberry PI. Thus, the present work consists of obtaining three-dimensional information through two webcams, for the representation of this a disparity map filtered by Weighted Least Squares is calculated and an algorithm based on the averages of three different regions is proposed. In this way, the system makes a decision which is transmitted to the System User through connected headphones. The system alerts to avoid collisions, with voice commands indicating to the user the best route to avoid obstacles. This system is implemented in a base-3D printer which has a compartment for a battery, which will feed the RaspBerry PI.