The architecture and recognition algorithm in Haibao perceptual development robot

Haibao, demonstrated at World Expo 2010, has dense LED lights all over its body for flexible changes of its facial expressions, clothes and display of other graphic information. Its visual capability is performed by a Where-What Network (WWN), modeling the dorsal and ventral streams of the biological visual cortex. It is adapted from WWN-3, the first series of visual detection and recognition algorithms that are capable of detecting and recognizing a general object or pattern from complex backgrounds. In this paper, key issues in the architecture, detection and recognition algorithm, and the robot body are discussed with some results of experiments.