An advanced visual tracking and stable grasping algorithm for a moving object is proposed. The stable grasping points for a moving 2D polygonal object are obtained through the visual tracking system with the Kalman filter and image prediction technique. The accuracy and efficiency are improved more than any other prediction algorithms for the tracking of an object. In the processing of a visual tracking. the shape predictors construct the parameterized family and grasp planner find the grasping points of unknown object through the geometric properties of the parameterized family. This algorithm conducts a process of ‘stable grasping and real time tracking’.