Jin Jianan

St. Thomas Aquinas Regional Secondary School, North Vancouver, Canada

Abstract：

With the increasing demands for intelligent manufacturing and precision assembly, CNC robot assembly places higher requirements on the real-time performance, accuracy, and environmental adaptability of visual servo control. Traditional single-vision methods have perceptual limitations in dynamic and complex scenarios, making it difficult to meet the needs of high-precision tasks. This study systematically reviews the evolution and industrial applications of YOLO target detection technology, the principles and accelerated development of NeRF neural radiation field technology, and the theoretical practice of visual servo control in robot assembly. Based on this, the design principles and implementation paths of the YOLO and NeRF fusion framework are discussed, a visual servo optimization strategy based on semantic detection and 3D reconstruction is proposed, and its adaptability to the constraints of CNC robot assembly scenarios is analyzed. Research shows that this fusion method can effectively combine real-time 2D semantic information with high-fidelity 3D geometric representation, improving pose estimation accuracy and control robustness, providing a more reliable perceptual foundation for precision assembly. Future research will further optimize computational efficiency and generalization ability, and promote applications in the field of intelligent manufacturing through experimental verification.

Key Words：

YOLO object detection; nerf neural radiation field; visual servo control; CNC robot assembly; fusion framework