Deep Coding Patterns Design for Compressive Near-Infrared Spectral Classification

TL;DR

This paper introduces an end-to-end method for jointly designing coding patterns and neural network parameters to improve spectral classification accuracy directly from compressive near-infrared measurements, outperforming traditional methods.

Contribution

It presents a novel joint design approach for coding patterns and neural networks in compressive spectral imaging, enhancing classification accuracy without reconstruction.

Findings

Outperforms traditional and random coding designs by up to 10% in accuracy.

Validated through extensive simulations on the 3D-CASSI system.

Demonstrates the effectiveness of end-to-end optimization in spectral classification.

Abstract

Compressive spectral imaging (CSI) has emerged as an attractive compression and sensing technique, primarily to sense spectral regions where traditional systems result in highly costly such as in the near-infrared spectrum. Recently, it has been shown that spectral classification can be performed directly in the compressive domain, considering the amount of spectral information embedded in the measurements, skipping the reconstruction step. Consequently, the classification quality directly depends on the set of coding patterns employed in the sensing step. Therefore, this work proposes an end-to-end approach to jointly design the coding patterns used in CSI and the network parameters to perform spectral classification directly from the embedded near-infrared compressive measurements. Extensive simulation on the three-dimensional coded aperture snapshot spectral imaging (3D-CASSI) system validates that the proposed design outperforms traditional and random design in up to 10% of classification accuracy.