Ensemble of SVM for Accurate Traffic Sign Detection and Recognition

Abstract

Automatic traffic sign detection and recognition plays a very significant role in advance driver assistance system and intelligent transportation system. In this paper, approach for circular traffic sign detection and recognition is proposed. The entire performance of the proposed system is calculated on German Traffic Sign Detection Benchmark (GTSDB) and German Traffic Sign Recognition Benchmark (GTSRB) datasets. Traffic signs are detected on color images based on RGB color thresholding technique and further detecting circle using circular Hough Transform. In traffic sign recognition, features are extracted using Histogram of Oriented Gradient (HOG) and strong components of the image are selected by Principal Component Analysis (PCA) and classified using Ensemble of SVM as the size of the dataset is increasing day by day. Results obtained undergoes statistical test showing the better performance of the algorithm proposed.