Recognition of road signs from video
Paweł Pawłowski, D. Prószyński, Adam Dąbrowski · Signal Processing: Algorithms, Architectures, Arrangements, and Applications · 2008
In this paper we propose an algorithm for real time detection and recognition of main groups of vertical road signs. The described recognition process consists of three following stages. In the first we check, which objects in individual images of the examined video have similar colors to those of particular road signs. To solve this problem, i.e., to determine proper RGB dependencies for the color-based segmentation of real world scenes, we use perceptive color spaces such as HSI and JCh instead of the RGB space. The next step is a test to determine which regions from those just found in the previous stage by the color matches exhibit the road sign like shapes. This classification is done by computing geometric moments and appropriate shape factors. As the meaning of most road signs lies in the shape of the symbol just inside the sign, the essential process consists in recognition of this symbol. Thus in the third and final stage we recognize and classify the sign symbols using an artificial neural network (ANN).