Shape-Based Traffic Sign Recognition Using Biologically Motivated Features

Abstract

This paper presents a shape-based biologically motivated research study to recognize the Iranian traffic signs. A biological model called HMAX is used as feature extractor to deal with traffic sign images, taken in real scenes. The extracted feature vectors are simply classified using a K-Nearest Neighbor classifier. Proposed model is implemented using a self collected dataset to evaluate its effectiveness. Three experiments have been implemented to evaluate performance of the proposed model, compared to some conventional models such as PCA, DCT and 2DPCA. The experiments have been done in cases in which the position, scale and the viewing-angle of objects in images are variable. The difference in recognition rate of the proposed model in comparison with other forging methods was impressive in all implementations. In each experiment the difference in performance of proposed model compared to conventional models was about 60% on the same database. The high model’s recognition rate will increase system stability and reliability on real time application.