Digital twin-trained deep convolutional neural networks for fringe analysis

High-speed three-dimensional (3D) fringe projection profilometry (FPP) is widely used in many fields. Researchers have recently successfully tested the feasibility of performing fringe analysis using deep convolutional neural networks (CNN). However, the existing methods require tremendous real-world scanning trials for model training, which is not trivial. In this work, we propose a framework to establish the digital twin of a real-world system in a virtual environment and a process to generate 3D training data automatically. Experiments are conducted to demonstrated that a physical system could adopt the CNN trained in the virtual environment to perform accurate real-world 3D shape measurements.

Automated summary

The study proposes a framework to establish the digital twin of a real-world system in a virtual environment and a process to generate 3D training data automatically. Experiments demonstrate that a physical system can adopt the CNN trained in the virtual environment to perform accurate real-world 3D shape measurements.