Deep Learning Enabled Real Time Speckle Recognition and Hyperspectral Imaging using a Multimode Fiber Array

TL;DR

This paper presents a deep learning approach for real-time hyperspectral imaging and speckle pattern analysis using a multimode fiber array, outperforming traditional methods in speed and robustness.

Contribution

It introduces a neural network-based method for spectral deconstruction from speckle patterns, enabling fast, robust, and real-time hyperspectral imaging with a fiber array.

Findings

Deep learning achieves reliable spectral reconstruction from speckle patterns.

The method outperforms analytical and compressive sensing techniques in speed and noise robustness.

Real-time hyperspectral imaging demonstrated with a fiber array system.

Abstract

We demonstrate the use of deep learning for fast spectral deconstruction of speckle patterns. The artificial neural network can be effectively trained using numerically constructed multispectral datasets taken from a measured spectral transmission matrix. Optimized neural networks trained on these datasets achieve reliable reconstruction of both discrete and continuous spectra from a monochromatic camera image. Deep learning is compared to analytical inversion methods as well as to a compressive sensing algorithm and shows favourable characteristics both in the oversampling and in the sparse undersampling (compressive) regimes. The deep learning approach offers significant advantages in robustness to drift or noise and in reconstruction speed. In a proof-of-principle demonstrator we achieve real time recovery of hyperspectral information using a multi-core, multi-mode fiber array as a random scattering medium.