Closed fringe pattern demodulation using deep learning

• Main Author: Ong, Zoey
• Other Authors: Qian Kemao
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/172136
• Institution: Nanyang Technological University

Closed fringe patterns present a complicated challenge in phase demodulation and is considered the most difficult type of fringe pattern to be processed. Among existing phase demodulation techniques, there still exists a gap whereby popular traditional techniques are unable to extract phase information with high accuracy from single-frame fringe patterns with closed fringes. Deep learning shows promise in phase demodulation, and have proved to be useful in other phase related fields such as phase unwrapping. Deep learning models could be able to identify complex relationships found between closed fringe patterns and their corresponding wrapped phase maps, which could fill the gaps of current methods. Using a neural network model closely inspired by the U-Net model architecture, this paper aims to demonstrate the potential of deep learning in extracting intricate phase information from closed fringe patterns. The U-Net model excels in capturing spatial information and features, which comes in useful when processing image data of fringes. Experimental results show that the proposed model is able to achieve a relative RMSE of around 4% based on stimulated closed fringe patterns.