An illumination adaptive color object recognition method in robot soccer match

Abstract

This paper presents an illumination adaptive color object recognition method for robot soccer match. Generally, the colors are identified by referring to the pre-defined bounds for the components of each color. However, it is not an easy work to define non-interfered bounds for different colors, and the bounds are sensitive to illumination conditions. Instead, in this work, different colors are discriminated by comparing their chrominance component in the YUV color space. With the combination of the geometry properties of the color labels of the objects, the recognition process consists of two stages. In each stage, the color labels' recognition is realized by simply comparing the average chrominance component of the separated regions. Thus the system can be adaptive to the variation of ambient illumination. Experiments prove this method's validity and reliability.