The PhD fellowship is centered around the exploration and development of digital twins in manufacturing systems, virtual models that mirror physical systems for real-time analysis, monitoring, and optimization. These digital twins are instrumental in modern smart industries, aiding in production planning and control by merging past, present, and future data for intelligent decision-making. Specific applications include real-time monitoring and production control, yet their creation and maintenance present challenges such as data consistency and timely input.
This research project aims to tackle these challenges by incorporating computer vision to facilitate Automated Digital Twins Generation. Using cameras and sensors, the work will focus on capturing physical asset data, addressing complex issues such as 2D object recognition, 3D volume estimation, and multiple view integration. The outcome is an application for camera-based digital twin production systems involving human operators. By employing deep learning models and providing a flexible, real-time approach, this work seeks to reduce the time and cost of constructing digital twins while adapting to various manufacturing scenarios.