Automation of Part Detection and Gripping Using Rectangle Detection and Surface Anomaly Detection Inside Holes Using an Autoencoder.

During the drilling of holes in aircraft hydraulic components, burrs form at the intersections of the holes. Burrs cause stress concentration and flow obstruction, so deburring and chamfering are necessary. Currently, these tasks are performed manually by skilled workers. To reduce the workload and improve production efficiency, full automation of this process is required. This paper focuses on detecting metal parts and evaluating the surface characteristics of the inner surfaces of the component holes. High-accuracy part detection was achieved using keypoint matching and other techniques. Additionally, a system was developed using an autoencoder to determine the presence of anomalies on the inner surfaces of the parts. Future work should establish methods capable of identifying the position and size of burrs.