Abstract: To address the issue of low work efficiency caused by the reliance on manual processes in the mold clamping phase of riveting, this paper presents a design of an end effector for a clamping robot to enhance automation and operational efficiency.To overcome the challenge of workpieces extending beyond the camera’s field of view, an uncalibrated vision servo controller based on improved image moments is proposed.During the clamping process, the recursive least squares（RLS） method is employed to estimate the image Jacobian matrix online, while incorporating an extended Kalman filter to provide real-time feedback on the joint movement of the robotic arm to the RLS estimation module.Simulation results demonstrate that this control method effectively extracts image features both within and beyond the camera’s view.Compared to the sole use of the RLS method, the image moment-based control approach exhibits superior performance in feature extraction efficiency and noise resistance, with a faster convergence rate for the image moment feature error.Physical experiments further validate the accuracy of the method when workpieces are outside the camera’s field of view, successfully addressing the limitations of traditional methods.This advancement enables the clamping robot to accurately locate workpieces and reach target positions, thereby significantly improving the automation level and work efficiency of the riveting process.