Utilização da visão computacional e detecção de características para auxiliar na navegação de pessoas com deficiência visual em ambientes internos

Abstract

Visually impaired people face many obstacles in a daily basis. Computer vision can be used to improve their quality of life, contributing to enhance the accessibility of places. This work proposes the use of computer vision and image recognition to aid indoors navigation of visually impaired people. A prof-of-concept system was developed using a mobile application and a web server containing a set images and metadata describing the environment. The mobile application in possession of the user captures images and gives audio feedback based on the recognition of that image by the server. The image recognition and audio feedback are based on images and their metadata previously added to the server. The main contribution of this work is the implementation of a system able to help visually impaired people, along with a literature review about image manipulation, feature extraction and image recognition. A set of tests were performed to evaulate the system accuracy, aiming to simulate real use cases such as when an user taking a picture stands in front of or besides an object.