Sistema de detección de señales de tráfico para la localización de intersecciones viales y frenado anticipado

Abstract

This paper describes a system for traffic sign detection from surrounding traffic intersections and analysis for the ability of detection depending on the distance is presented. The method is based on segmentation by color on the RGB-normalized space (ErEgEb) for generating regions of interest (ROIs), the classification of the type of signal using YCbCr and ErEgEb spaces to build a statistical template, where to remove the background has proposed a probability distribution function that models the color of the objects of interest and its background. This system is specialized in Stop and Give way signs.After that when compared to the Viola & Jones method where the detection capability is verified at different distances is then analyzed. For distances less than 48 meters, this method reaches a detection rate of 87.5%, in the case of Give way sign and 95.4% regarding the Stop sign, while for distances less than 30 meters the detection rate is 100%. These results are higher than those in the state of the art. Experiments have been developed over a traffic signs database, generated from images taken in several streets of the Santiago, Metropolitan Region, Chile, using an experimental vehicle designed to develop intelligent systems.