An approach to a fuzzy color detection method, robust with regard to variable illuminant level conditions

Abstract

We propose a method for characterizing a set of training colors, based on sample color pixels captured with a camera. The system tries to detect the training colors on test pixels within images captured under illumination level conditions different from those used in the training process. The training color characterization is based on Smith's perceptual color model, but using only hue and saturation components. Both, training and classification processes make use of fuzzy techniques to assume vagueness involved within training and test data.