A SIFT Method Application for a Six DOF Manipulator Grasping an Object against Complicated Environment

This paper presents an image process based results for controlling a manipulator grasping an object against complicated background environment. Some problems caused by environment which exist including 1) orientation and distance measurement inaccuracy - 2) image distortion from lens - 3) grasp point estimation and force-closure judgment - 4) pose control of manipulator - 5) complex background and light reflection. The work begins from perception of environment and objects by image system. The specified objects searching based on SIFT approach keeps working with the pan-tilt vision till finding out those targets. The information at least provided by SIFT^reg contains 1) recognition confidence (%) - 2) centred location (X, Y) of an object in image frame and estimated distance in absolute frame (D) and 3) matched index (user defined text) for an object. By these data, the manipulator controller can align the gripper frame to the object as close as possible and execute grasp mission. Both searching and tracking for specific objects in these configurations are proven powerful.