Automatic Rebar Binding Method for PC Components based on RGB-D Information

Aiming at the problem of the low efficiency of manual steel bar binding in precast concrete components, this paper proposes an automatic steel bar binding system for PC components that integrates RGB-D information. By improving the YOLOX-s network to detect the status of binding points, extract the area of unbound points, get the skeleton of the image, combine the image point cloud information, and the hand–eye position relationship to locate the three-dimensional position information of the binding points. The RRT-connect algorithm plans the binding trajectory, and the robotic arm executes the process. The model’s mAP is 97.91%, and the FPS is 38.76[Formula: see text]fps. The results show that the average time to locate the binding point is about 108[Formula: see text]ms, and the average time for completing a binding is about 15[Formula: see text]s, meeting real-time binding requirements.