Adaptive color claddification with gaussian mixture model

In this paper we present an adaptive color classification algorithm for RoboCup. Color-based vision faces the challenge of perceived color being variant to illumination. We propose a color classification algorithm that is robust and reliable under dynamic lighting conditions. Supported by the dichromatic reflectance model, we use a Gaussian mixture model (GMM) of two components to represent the distribution of a color class of interest in the YUV space. The color model is continuously updated to achieve adaptation. We show experimentally that a GMM with two components can be used as an accurate and complete representation of a dichromatic surface, and that our algorithm is capable of adapting and classifying color classes in real time