A visual-data processing method for an image feedback control of industrial robot with a hand-held camera is proposed, in which the size of object and the distance between robot gripper and object are calculated in real-time without any a priori information and used for guiding the robot at each frame time as the robot approaches to the object.
The procedure can be summarized as followings; Firstly, the image features are obtained from boundary tracking method and by cross-template shape in its interior. Secondly, the image is tracked by cross-template matching algorithm without boundary information as the image is moved at each frame time. Finally, the distance between robot gripper and object and its real size information of object is calculated by the image shape-deforming and camera motion.
The proposed method is implemented and tested to the PUMA 560 industrial robot with a newly designed hand-held camera system for tracking an arbitrarily positioned object in work space.