摘要
This study presents a novel strategy for improving robotic hand grasping via Mediapipe-based virtual contour method. The method allows the adjustment of hand gestures based, calibrates using a grasp matrix, and mitigates common issues in AI-driven grasping. Effective finger posture correction and stable force distribution for object grasping of rectangular objects with different thumb position have been demonstrated with this method through calibrated experiments. The proposed method is well-suited for scenarios where a robotic hand is directly controlled using human hand posture.