Research Of Fast Recognition Of Robot-Soccer Based On Color Attributes

Aizeng Cao,Jinping Li,Baohua Zhao

2006-01-01

Abstract:According to the real-time requirement in the Robot-Soccer game, the paper, firstly, presents a color classification method, which combines a group of color classifiers obtained through Artificial Neural Net training with the threshold vector method in the YUV color space, and farther presents a new searching and locating algorithm of the object robot based on the theory of the feature point filling, that is, delimiting the object through bisection searching and precisely recognizing and locating the little vehicles with the statistic of pixels of the team mark's and the team member mark's different colors in the neighborhood of the center point. This algorithm can obviously reduce the calculating amount in the process of recognition. The validity of the algorithm has been proved in the experimentation of the RoboCup. It can faster recognize the object, guaranteeing the precision. And it has a faster recognition speed compared with other searching algorithms.