Abstract: To obtain a higher recognition rate and a better robustness to a change in the intensity of illumination, this paper presents an approach of color recognition based on a combination of SVM classifier and Boosting classifier (SVM-Boosting classifier) for the color identification of ping-pongs. First, two color features independent on the intensity of illumination are selected to train SVM classifier, and four color features are selected to generate Boosting classifier. Finally, the SVM-Boosting classifier is constructed. The results show that the method gives a very higher recognition rate, up to 99.4%, for ping-pongs recognition, a higher recognition rate compared to the threshold method and SVM classifier, and a faster recognition speed compared to Boosting classifier. The method can satisfy both sides the recognition accuracy and speed, and is particularly suitable to a visual system with high accuracy and large sample capacity.

Key words: color recognition, classifier, visual system