Traffic Signs Recognition Based on Color and Shape Information

摘要

Traffic signs recognition has received an increasing interest in the last years. This paper proposes a method for circular traffic signs recognition using a novel system composed of two cameras and two galvanometer mirrors. The detection and classification stages are processed in captured images by wide and narrow angle cameras separately. The contour shape information is used for detecting the candidate positions of traffic signs in wide angle camera. Galvanometer mirrors are controlled according to centers of candidates. High resolution images of candidates are captured using narrow angle camera and used for classification. In this step, ellipse detection method is applied to get the exact region. Segmented result will be matched with the templates to realize the classification.