Sistema de Reconocimiento de Señalamientos en Entornos Abiertos para la Orientación de Personas con Discapacidad Visual

This work presents the development of a prototype that allows identifying specific signs through artificial vision techniques. It uses an image segmentation stage based on the SLIC superpixel algorithm, followed by a sign recognition and classification stage based on convolutional neural networks and has been implemented in a simple-board computer (SBC). The prototype informs the user of the identification of these signs through an audio message sent to headphones, it has a GPS module that obtains the location where the sign was recognized and is stored to offer the user notifications about nearby signs. The tests were performed with 1: 2 scale signs in open spaces, with natural light. The prototype is intended as a support for visually impaired people to move in open urban environments. Processing times and prototype performance are reported. Although the implementation in the selected simple-board computer makes its use unfeasible due to operating times, the functionality of the system is demonstrated.