Deep learning method for optical fiber curvature measurements based on time series data
Brian Pamukti, Muhammad Fajar Faliasthiunus Pradipta, Shien‐Kuei Liaw, Fu-Liang Yang, Yamei Yang · Journal of the Optical Society of America B · 2024
Curvature detection is an essential technique for monitoring landslides, which are frequent and destructive disasters. Existing methods for curvature detection using fiber-optic sensors have limitations such as complex fabrication or large data size. We propose a data processing method for high-accuracy curvature detection that employs deep learning. We experimented using different levels of curvature and compared our method with other methods. Our method achieves 99.82% accuracy for classification and root mean square error of 0.042m −1 for regression with a simpler structure and smaller data size. Our approach demonstrates its potential for landslide detection and integration with communication systems.