A fast Raffic sign recognition algorithm for gray value images

Abstract

A fast shape recognition method is presented for detecting traffic signs in monochrome images. Based on Marr’s primal sketch we used a hierarchical grouping method to link low-level features of pixels like gradient orientations to high-level features like triangles and ellipses. The method is fast, because it is based on local pixel operations instead of analyzing the whole shape. Examining the local area around a pixel for parts of geometrical shapes, results are symbolic representation which are linked back to the pixel, i.e. the pixel is part of the geometrical shape. A unique description of the geometrical shape is calculated from pixels lying on the shape.