Recognition of Mandatory Traffic Signs using the Hausdorff distance

This paper proposes a new recognition algorithm of mandatory traffic sings using the Hausdorff distance. This algorithm has been designed to detect arrows on traffic signs, especially direction signs. The arrows on these mandatory signs appear in a multitude of different forms and positions. Due to this variety this algorithm uses a structural approach to recognize the arrows. First the sign is transformed into a binary bitmap. The second stage consists in computing the skeleton of the arrows that appear in the mandatory signs of the image. The algorithm calculates de Hausdorff distance between some models of skeletons of arrows and the skeleton obtained. The algorithm recognizes the model that produces the smallest Hausdorff distance. Key-Words: ideogram, arrow, mandatory, traffic signs, recognition, Hausdorff distance, Voronoi surface, skeleton.