Detection and recognition of road signs in a video stream based on the shape of the panels

In this paper, we present a robust approach of automatic detection and recognition of road signs in national roads, starting from the images resulting from a video stream taken by a camera embarked on a vehicle. Our approach is composed of three main phases: the first phase is to extract video stream images containing a circle or a triangle. This extraction is performed respectively by Hough transformation and Ramer-Douglas-Peucker filter, the second phase consists of extraction areas of the calculated image, in the previous phase. In the third and last phase, we proceed to a matching of the extracted image areas with signs of reference by comparison of interest points extracted by the SURF method and the matching method FLANN.