A comparative study of classification methods for traffic signs recognition

Abstract

This paper presents a comparative study of several classification methods for the task of recognizing traffic signs in urban areas. These classification methods are artificial neural network (ANN), k-nearest neighbors (kNN), support vector machine (SVM), and random forest (RF). First, HSI-based color segmentation process is applied to obtain candidate regions. Using centroid-based feature, these regions will be classified into three shape classes, such as circle, rectangle and triangle. Hereafter, histograms of oriented gradient (HOG) features are extracted from each region that will be utilized in recognizing step. For comparison, well-known public databases will be used. The comparison based on the implementation result from those data with difference condition of intensity and angle of view. Comprehensive comparative results to illustrate the performance result of each classification method are presented.