Abstract:This paper describes a visual tracking method which uses image moment as image feature for tracking a target that moves in 3D with translational motion. According to the specific tasks, a set of moment is selected as image feature. Then, the moment-based Jacobian is deduced. With the moment-based Jacobian, the visual servoing controller composed of image-based feedback and adaptive motion compensation is designed to track the moving target. In the tracking process, the knowledge of camera focus and depth of object is unnecessary. The simulation results show that using image moment as image feature can avoid the complex feature matching process, and acquire a satisfied tracking accuracy.
1 Feddema J T, Lee C S G, Mitchell O R. Weighted Selection of Image Features for R esolved Rate Visual Feedback Control, IEEE Trans on Robotics and Automation, 1991,7(1): 31-47 2 Hashimoto K, Kimoto T, Ebine T, Kimura H. Manipulator Control with Image-Bas ed Visual Servo. in Proc. of the 1991 IEEE Int′l Conf on Robotics and Automat ion, Sacramento, California, April 1991, 2267-2272 3 Papanikolopoulos N P, Khosla K P. Feature Based Robotic Visual Tracking of 3-D Translational Motion. in Proc of the 30th on Decision and Control, 1991.18 77-1882 4 Wells G, Venaille C, Torras C. Promising Research Vision-based Robot Positio ning Using Neural Networks. Image and Vision Computing, 1996,14(10) : 715-732 5 Martinez J, Thomas F. A Reformation of Gray-Level Image Geometric Moment Com putation for Real-time Application. Proc of IEEE Int, Conf on Robotics and Aut omation, 1996. 2315-2320 6 Prokop R J, Reeves A P. A Survey of Moment-based Object Representation and R ecognition, CVGIP: Graphical Models and Image Processing, 1992,54( 5): 438-460 7 李介谷.计算机视觉的理论和应用.上海交通大学出版社 1991