Modelo de detecção e reconhecimento de semáforos baseado em Atenção Visual e Inteligência Artificial

Abstract

It its known that there is a large amount of visual information present in the traffic, which increases the possibility of traffic signals go undetected, which could cause a serious accident. Besides this, there is a increasing the number of research related to intelligent transport systems, in which vehicles can be self-guided, probably having to coexist with manually guided vehicles. Watching these two facts, a detector and recognizer of traffic lights becomes very useful. This course conclusion work aims to propose a detection mechanism and recognition of traffic lights based on biological concepts of artificial intelligence, specifically visual attention, and image processing. The proposed mechanism uses information from the color histogram of the detected area for classifying scene as having a red or green light, and achieved good functioning in the experiments presented in the work. The experiments were performed in the daytime and nighttime periods.