ABSTRACT
In this chapter, the design of an on-board processor which enables recognition of a given road sign from a vehicle in motion is presented. A safety system to be installed in vehicles could be based on this processor in order to automatically detect and identify road signs. Afterward, the recognition system could make an objective decision according to the information detected. One of the greatest difficulties in achieving this goal lies in the number of different distortions that may simultaneously modify the reference sign. Variations in scale, in-plane and out-of-plane rotations, background clutter, partially occluded signs, and variable illumination are some examples of distortions that can affect road signs. To overcome these problems, a number of techniques have been studied in pattern recognition [1-13]. Some of them have been applied to road sign recognition. For instance, an optical correlator for scale-invariant road sign detection was proposed in Refs. 14 and 15. Recently, partial tolerance to in-plane rotations and scale invariance has been obtained by using partially invariant filters in a multiple correlator [16,17].
