Traffic sign recognition by Bags of features

Road sign recognition method has been studied for realizing drivers assisting system, and various methods have been developed. Road sign alters its shape and color depending on its relative location against camera and surrounding condition such as weather and daytime. The object detection method utilized Scale Invariant Feature Transform (SIFT) is used in this study. Unless its advantage in the robustness property, calculation costs both for detecting SIFT and matching with database are usually expensive. In this paper, region extraction is performed for reducing SIFT detection cost, and the Bags of Features method is applied for traffic sign recognition. At the recognition phase, support vector machine (SVM) approach is used. Through several experimental results, lower calculation cost and higher accuracy rate (97.5 %) are observed.