Desenvolvimento de um sistema de reconhecimento de sinais de trânsito utilizando processamento de imagens com OpenCV para um robô humanóide.

Abstract

This dissertation presents a recognition method of pedestrian crossing, coupled with the necessary theoretical bases and concepts required to develop a computer vision system to recognize patterns and identify real-word objects. Functions of the OpenCV (Intel Open Source Computer Vision Library) image-processing library are used to develop this system, which was designed to run on a DARwIn-OP (Dynamic Anthropomorphic Robot with Intelligence - Open Platform) humanoid robot. Tests were performed in a real traffic environment, in different conditions of visibility and brightness, for system evaluation and validation. Results show how efficient OpenCV functions are in this kind of system, which requires a fast response time on hardware with reduced capabilities, especially memory and processing power.