A Dynamic Active Contour Model for Visual Tracking of Robotic Manipulator Motion Trajectory

Abstract

Visual tracking of motion trajectory is a typical issue in visual servo control of robotic manipulator. A dynamic active contour model is proposed to track robotic motion in captured image sequences. It converges at the trajectory distribution to get minimum energy of snake functional. The manipulator can be localized through searching the converged energy points set in spatio-temporal space. Energy weights in the function are real-time adjusted to avoid local minima during convergence. To improve snake searching efficiency, quadratic trajectory least square estimator is utilized to predict manipulator motion position before tracking. A fast visual tracking algorithm was developed integrated with the dynamic snake model and the least square trajectory estimator. Experimental results of micromanipulator motion tracking demonstrate the performance of the proposed approach.