Vision-Based Grasping Method for Prosthetic Hands via Geometry and Symmetry Axis Recognition.

This paper proposes a grasping method for prosthetic hands based on object geometry and symmetry axis. The method utilizes computer vision to extract the geometric shape, spatial position, and symmetry axis of target objects and selects appropriate grasping modes and postures through the extracted features. First, grasping patterns are classified based on the analysis of hand-grasping movements. A mapping relationship between object geometry and grasp patterns is established. Then, target object images are captured using binocular depth cameras, and the YOLO algorithm is employed for object detection. The SIFT algorithm is applied to extract the object's symmetry axis, thereby determining the optimal grasp point and initial hand posture. An experimental platform is built based on a seven-degree-of-freedom (7-DoF) robotic arm and a multi-mode prosthetic hand to conduct grasping experiments on objects with different characteristics. Experimental results demonstrate that the proposed method achieves high accuracy and real-time performance in recognizing object geometric features. The system can automatically match appropriate grasp modes according to object features, improving grasp stability and success rate.