A Simple Screw-Hole Discrimination Pipeline for Deployment in Autonomous Manufacturing

Autonomy in manufacturing is a major challenge of Industry 4.0. To allow autonomous assembly of rotary tables, one important aspect is the localization and discrimination of empty holes and screws. In this paper, we propose a novel feature descriptor to discriminate screws and holes. The descriptor is embedded in an end-to-end pipeline, which localizes candidates on concentric workpieces. The candidates are analyzed using a slim tree-based classifier, which discriminates inserted screws and empty holes. These are later used in autonomous screwing processes. The proposed methods are supplemented by an intuitive UI-driven process, which enables non-expert users to train and deploy the them leniently in autonomous manufacturing processes. The feasibility of the proposed method is demonstrated in multiple lab experiments and in a real-world demonstrator as part of the SHAREWORK project.