A visual automatic objective cognition method using human instruction for robot imitation

A visual automatic objective cognition method using human instruction for robot imitation learning is proposed. The image sequences of human instruction are segmented into some sets of image s sequences of object using dynamic detection. These images are extracted from some features in the defined features pool to construct a features database. Features in this database are clustered based on density to refine clusters of conceptions corresponded to objects. This method can cognize the objects in daily environment autonomously and incrementally with the help of human's instruction, and the results of experiment prove effectiveness of this method.