Recognition of Traffic Signs Based on Color Features and Neural Network Model

This paper presents an automatic traffic sign recognition system based on color, shape features and neural network model. The system consists of two parts: detection and recognition. With dominant color and shape features, traffic signs are detected using color and shape information. Firstly, HSI space being immune to illumination change is utilized to detect candidate traffic signs. Then an algorithm based on shape feature is proposed to get correct signs region from candidate traffic signs. Once the sign has been detected the recognition is done. Sign feature is extracted using APEXNN, and SOMNN is utilized to recognize an object in a determined category of objects. The training set with noise, scale, rotation and distortions is created to train the nets. The experimental results show the feasibility and robustness of the proposed algorithm.