Real time trajectory matching and outlier detection for assembly operator trajectories

Flexible, reactive and adaptive manufacturing systems are a prerequisite to cope with the demand for low volumes of highly customized products of today’s market. For years, manufacturing companies have been using real-time data capturing systems, such as RFID, to gather the necessary data to obtain insights in their production processes, mainly in the domain of quality control and inventory management. However, very few work has been done on monitoring an assembly operator during his work cycle in real-time. This paper presents a method to match operator trajectories, obtained through a multi-camera vision system, in real-time to predefined models. This way, the performance of the operator can be assessed online and problematic or anomalous work cycles can be detected. This information can then be used to support the operator in his pursuit for continuous improvement by pointing out improvement potential.