Fast Road Signs Recognition Using Contour Features

This paper proposes a fast road signs recognition framework with shape filtering and contour features. We observe the standard road signs and construct two basic feature tables for recognition processing. In road driving images, we transfer the RGB color domain to HSV for the red regions detection firstly. And we obtain candidate regions and decide the kinds of shapes by eliminating the redundant regions with the ratio of width and height of bounding boxes and predefined shape filtering. For the recognition of road signs, we calculate the feature values of contour in 36 directions. And feature matching and voting process are utilized to recognize the content of road signs. Our method is fast and efficient. Furthermore, the proposed method could be embedded in real-time visual vehicle driving assistant systems.