Recognition and Tracking of Metal Clip-like Objects Based on Deep Learning Semantic Segmentation

The problem of feature recognition and servo tracking of semantic segmentation results for metal gripper jaws is investigated. In this paper, an algorithm based on the combination of deep learning and visual servo is proposed to design a servo control law based on the parameters of the size of the operating area and servo speed, so as to control the movement of the robotic arm in the direction of the three rotational degrees of freedom in the tool coordinate system. In this paper, we built a tracking experiment platform, wrote a control algorithm and a robotic arm control program. After that, tracking experiments were carried out by combining semantic segmentation and feature extraction algorithms, and the feasibility of the method was finally verified.