Image processing algorithms running on real-time over an embedded system

Abstract

This document and project has the purpose of developing a system able to help blind people orient themselves using computer vision techniques. Computer vision is becoming increasingly essential to our everyday lives; it is heavily used across the automation of industries, medical imaging, as well as self-driving vehicles such as cars and drones. When associated to artificial intelligence, computer vision techniques enable the algorithms to do wonders, as a matter of facts it is used this way in most cases nowadays. The project’s goal is to grab images from a helmet mounted on the head of a blind person, composed of a camera, an embedded computer and an audio headset, those images are then analyzed on the embedded computer in real time to make guesses on the position of the followed person or object in that image, the position is then described to the helmet wearer by a sound. The embedded computer is equipped with a remote desktop server and therefore a graphical user interface has been made to choose and set parameters for the computer vision software, such as which algorithm to use and what object or person to follow.