Bolt identification and positioning method based on YOLOv5

Prefabricated steel structure buildings are increasingly being applied in the construction of transmission towers, substation infrastructure projects, and main power plant buildings. Bolt is an important connection method in the connection process, and bolt tightening is an engineering problem that needs to be solved at present. There is an urgent need for the development of a bolt tightening robot specifically designed for steel structure connection nodes to address this issue. This paper focuses on the visual recognition and positioning technologies for such a robot, particularly the development and training of a deep learning model for bolt recognition, the acquisition of the bolts' three-dimensional coordinates within the camera coordinate system, and the real-time detection and feedback of coordinate information. The experimental results demonstrate the effectiveness of the proposed strategy.