AI Recognition Based on Multi-Target Images for Assembling Parts using Robot with Screwdriver and Vision System

The process of assembling parts using screwdrivers is one of the standard core processes in production industries. Currently, the automated assembly technology uses robotic arms combined with a screwdriver to lock the parts automatically by preset positioning. Using a robust CNN algorithm of image processing, we positioned the robotic arm with a screwdriver accurately and automatically. We developed a system using a 6-degree-of-freedom serial robot with a webcam camera and an intelligent automatic screwdriver for a more flexible and efficient method of assembling parts. The robot's coordinate system was calculated with the 20-point calibration function from the robot's controller to obtain the lowest error value. As an image processing model, the developed YOLOv5m model was evaluated by assessing the accuracy of the detection performance from the fl-score based on the confusion matrix. The CNN algorithm classified and identified the screws and assemble holes with their positions.