Hand posture extraction for object manipulation of a humanoid robot

Hand posture extraction is important for object manipulation of a humanoid robot. This paper focuses on the problem of hand posture extraction for object manipulation of human robot BHR-2, and proposes a novel method based on blue mark to get the hand posture. Based on the attached marks, the hand is identified and segmented using a method based on multiple visual cues integration. The vertices of the mark are obtained by using K-L transform, Hough transform and Least Square Estimator (LSE) methods together. The world coordinates of these feature points and the postures of the hand are obtained through coordinate transform. The effectiveness of the proposed method has been illustrated by the experimental results.