Low cost hyperspectral imaging using deep learning based spectral reconstruction

Tools

Tschannerl, Julius and Ren, Jinchang and Marshall, Stephen (2018) Low cost hyperspectral imaging using deep learning based spectral reconstruction. In: Hyperspectral Imaging Applications (HSI) 2018, 2018-10-10 - 2018-10-11.

[thumbnail of Tschannerl-etal-HIS2018-Low-cost-hyperspectral-imaging-using-deep-learning-based]
Preview
 Text. Filename: Tschannerl_etal_HIS2018_Low_cost_hyperspectral_imaging_using_deep_learning_based.pdf
Accepted Author Manuscript
Download (374kB)| Preview

Abstract

The increasing number of applications of hyperspectral imaging results in a high demand for low cost, mobile devices. We propose a multispectral imaging (MSI) system based on time-multiplexed lighting using RGB Light Emitting Diodes (LED). We train a deep neural network that maps low dimensional multispectral input onto high dimensional hyperspectral (HSI) output that is collected with a HSI camera covering the range of 400 – 950 nm. Results on the 24 colour patches of the Macbeth colour checker chart show that with only five multispectral bands, a very accurate reconstruction of HSI data can be achieved.

ORCID iDs

Tschannerl, Julius ORCID logoORCID: https://orcid.org/0000-0002-4613-1693, Ren, Jinchang ORCID logoORCID: https://orcid.org/0000-0001-6116-3194 and Marshall, Stephen ORCID logoORCID: https://orcid.org/0000-0001-7079-5628;
CORE (COnnecting REpositories)