Abstract
In this paper stereo vision based tracking system for humanoid robot is proposed. The goal is to locate and track the end-effector in unknown and dynamic environments. A robust approach is developed by utilizing visual features. Identical visual features such as color, disparity and adaptive template matching are used to detect and track the end-effector robustly. An improved active search method is used in order to realize effective tracking of end-effector. Active search stereo vision system of humanoid robot is designed. Two computers are embedded in to the system to realize real time robust tracking of end-effector. One computer is used for processing visual information and other one is responsible for servo control of pantilt platform on which CCD cameras are mounted. The performance of the developed modular visual cues based tracking of end-effector is exploited through experiments.