This study explores the application of hyperspectral imaging (HSI) and machine learning for non-destructive analysis of peat, a crucial component in whisky production. Peat imparts distinctive flavors to whisky, but its extraction disrupts ecosystems and releases carbon, contributing to climate change.
The researchers used visible-near infrared (VNIR) and shortwave-infrared (SWIR) cameras to acquire hyperspectral data from 35 categories of peat samples from six different sources. They then applied radiometric calibration and adaptive thresholding to extract representative spectral samples.
The team assessed the feasibility of their approach through two tasks: peat sample grading and property estimation. For peat sample grading, they employed a Support Vector Machine (SVM) classifier and achieved impressive overall accuracies of up to 99.65% using SWIR data, outperforming VNIR data.
For property estimation, the researchers used the SVM model to predict total phenol levels, moisture content, and organic matter in both peat and condensate samples. The results were highly accurate, with R-squared values reaching 99.25% for total phenol prediction using SWIR data.
The study demonstrates the remarkable potential of HSI and machine learning for non-destructive, rapid, and precise assessment of peat quality and phenolic content. This technology can enable distilleries to optimize peat usage, maintain consistent flavors, and reduce their environmental footprint, aligning with the industry's need for greater sustainability.
To Another Language
from source content
arxiv.org
Ключові висновки, отримані з
by Yijun Yan,Ji... о arxiv.org 05-06-2024
https://arxiv.org/pdf/2405.02191.pdfГлибші Запити