Research of illumination invariant for color image in Mirosot of Robot Soccer

The color feature recognition of the object is used in Semi-autonomous robot soccer system in FIRA. However, the color feature of the object changes when the illumination varies in the environment during the competition. In this paper, different color models are analyzed and compared, such as normalized rgb model, HS model, normalized yuv model etc. They have different recognition performances of color feature. A new color model HSv´ based on dichromatic reflection model is proposed. Trace criterion which is constructed by within-class scatter matrix and between-class scatter matrix is adapted to evaluate the recognition performance of the color model. Experiments have been carried out in Mirosot(Micro Robot World Cup Soccer Tournament) which is a platform for semi-autonomous robot soccer. The results of the experiments show the new color model achieves a high recognition ratio for the object recognition.