Automatic road sign detection and recognition based on neural network

Road sign detection and recognition is an integral part of intelligent transportation systems. It increases protection by reminding the driver current condition route, such as notices, bans, limitations, other valuable driving information. This paper describes a novel system for automatic road signs, which achieved in two main steps. First, initial image pre-processed using DBSCAN clustering algorithm. The performed based on color information, generated clusters are segmented artificial neural networks (ANN) classifier. resulting ROIs then carried out their aspect ratio size to retain only significant ones. Then, shape-based classification ANN classifier HDSO feature detect circular, rectangular triangular shapes. Second, hybrid defined recognize detected from first step. involves combination so-called GLBP-Color extension classical gradient local binary patterns RGB space self-similarity feature. ANN, AdaBoost, support vector machine have been tested with introduced one selected it outperforms two. proposed method has outdoor scenes, collection common databases, well known traffic community (GTSRB, GTSDB, STS). results demonstrate effectiveness our when compared recent state-of-the-art methods.