Abstract:Multi-class sensing recognition is important in optical fiber distributed vibration systems, since accurate detections on the vibrations can significantly improve the performance of the systems in engineering applications. In this work, we have developed an ameliorated deep learning (DL) approach based on the serial fusion feature extraction model for multi-category sensing recognition in optical fiber distributed vibration sensing systems. Thus, the effective features can be extracted automatically and recognized accurately with the proposed approach. Firstly, we use a conversion algorithm to convert time-domain signals into the two-dimensional grayscale images. Then the proposed multi-class DL approach can be trained, in an end-to-end manner, to extract the convolutional features within each frame and the temporal features between frames. Finally, the extracted features are passed to a Softmax classifier to compute the final results. This proposed model is investigated to recognize seven common sensing patterns collected by a fiber-optic interferometer based vibration system in the field trial. Recognition accuracy and efficiency of the proposed DL approach have been greatly improved. Specifically, an average accuracy of 96.16% is achieved in the ten-fold-cross evaluation. The recognition response time per sample has been greatly shortened to 11 ms. The results have been compared to the performance of a conventional DL recognition model and a handcrafted feature extraction scheme. It is found that our proposed scheme can achieve significant improvement in accuracy and stability compared to reported models. Therefore, the proposed DL approach shows great application potential in the field of optical fiber distributed vibration sensing systems.

Deep Learning-Based Object Classification Through Multimode Fiber Via a CNN-architecture SpeckleNet.

Spatially Variant Deblur and Image Enhancement in a Single Multimode Fiber Imaged by Deep Learning

Deep-learned speckle pattern and its application to ghost imaging

Deep Learning for Position Fixing in the Micron Scale by Using Convolutional Neural Networks

Deep learning sheds new light on non-orthogonal optical multiplexing

Learning to transmit images through optical speckle of a multimode fiber with high fidelity

Multimode optical fiber transmission with a deep learning network

Reconstructing Images Through Multimode Fibers From the Up-Conversion Speckle Patterns via Deep Learning

DSCNet: lightweight and efficient self-supervised network via depthwise separable cross convolution blocks for speckle image matching

High performance few-mode fiber-based light field direction sensing system using deep convolutional neural network: fiber speckle demodulation network (FSDNET)

High-generalization deep sparse pattern reconstruction: feature extraction of speckles using self-attention armed convolutional neural networks

Optical Fiber Distributed Vibration Sensing Using Grayscale Image and Multi-Class Deep Learning Framework for Multi-Event Recognition

Machine learning meets Singular Optics: Speckle-based Structured light demultiplexing

Fiber-Optic Shape Sensing Using Neural Networks Operating on Multispecklegrams

Deep learning based optical curvature sensor through specklegram detection of multimode fiber

11 TeraFLOPs per second photonic convolutional accelerator for deep learning optical neural networks

A deep learning model enabled multi-event recognition for distributed optical fiber sensing

All-fiber high-speed image detection enabled by deep learning

Multimode Optical Fiber Specklegram Pressure Sensor Using Deep Learning

An S-CNN-based phase conjugation method in imaging through random media

Object Recognition Through Scattering Media Using Convolutional Neural Network