Visual object recognition for smart manipulation

In order to achieve fine operation of dexterous hand,a new method is proposed. The method consists of detecting a targeted object by Kinect,conducting background subtraction for the obtained depth on the basis of frame difference method to acquire the motion point of the targeted object and judging the targeted object by the TS fuzzy logic along with the features of the obtained targeted object. This paper uses BH8-280 for an object grasping experiment to verify the accuracy and rapidity of the method. In this experiment the pose,size and shape of the objects can be determined. The property categories of objects are judged by the T-S fuzzy logic system. It is seen that the dexterous hand can grasp the object accurately. The research of this method builds a foundation for fine operation of a dexterous hand.