TSRS - A NEW APPROACH FOR TRAFFIC SIGN RECOGNITION USING THE SIFT ALGORITHM

Abstract

This paper proposes a new approach for traffic sign recognition using images captured by a low-cost mapping system. The proposed approach applies the SIFT algorithm to extract keypoint features that are used to evaluate the correspondences between a road image containing one or more plates and the images of traffic signs (templates). The BBF algorithm was used to efficiently evaluate the correspondence between the SIFT features. Finally, we propose a new algorithm to filter only the pairs of keypoints (image-template) that are compatible as well as the orientation and positioning.