Machine Learning based Classification of Traffic Signs Images from a Robot-car

In this paper we analyzed the performance of some machine learning techniques in order to create a robust model for an application to a robot car traffic on a road model. The techniques evaluated support vector machines and convolutional neural networks. Several classifiers from these techniques were tested on 3000 images of traffic signs that were collected from an application environment under different lighting conditions. In addition, other images were collected outside the road model and others on the web for a better robustness analysis of the different classifiers.Our experimental results suggest that the Convolutional Neural Network (CNN) model is more accurate than that of the Support Vector Machine (SVM). But CNN has an implementation difficulty compared to SVM. In addition, the use of CNN seems to be a complex task due to the fairly high response time.