Model for the classification of damaged and vandalized Mexican traffic signs

Abstract

The automatic classification of traffic signs is a tool to support drivers to have a safer driving when they are behind the wheel and even to assist the semi-autonomous driving of a vehicle. The objective of this work is to present a model capable and classify vandalized traffic signs using convolutional neural networks (CNN). The methodology used is divided into four stages: 1) the collection of 1110 images of traffic signs is carried out, 2) a simulation of vandalism is performed on the images of traffic signs, 3) a convolutional neural network is created, and 4) the convolutional neural network is trained with the different data sets obtained. According to the results obtained from the experiments, it is concluded that the proposed model allows to recognize Mexican traffic signs with an accuracy of 97.67%. In addition, an experiment was carried out to determine the importance of color when classifying the images, the results show that the model is capable of classify different grayscale traffic signs based on the objects and symbols that contain the signs with an accuracy of 80.66%.