Aflatoxin contamination level estimation in food using reflectance multispectral imaging based system

H.m.p.s. Madushan, Pasindu Malshan, K.k. Abewickrama, Vijitha Herath, Roshan Godaliyadda, M. P. B. Ekanayake, A. D. S. N. P. Athukorala, M.d.h.m. Kumari, Darsha UdayangaJournal of Agriculture and Food Research 2024 · DOI

Abstract

This study presents a non-invasive, non-destructive, and rapid method of classifying and estimating aflatoxin levels in food samples using a multispectral imaging. A cost effective, mass deployable in-house built multispectral imaging system, suitable for onsite testing and operating in range from 365 nm to 940 nm, is utilized to capture the multispectral images of the aflatoxin contaminated bread samples. Spatial and spectral corrections are performed to mitigate the imperfections of the multispectral imaging system, followed by image preprocessing steps to mitigate the effect of random noise. Linear discriminant analysis is used to reduce the dimensionality of the feature space. Ensemble classification analysis utilizing classical machine learning classifiers yielded an accuracy of 90 % when classifying samples according to aflatoxin contamination levels. A functional relationship between the Bhattacharyya distance of noncontaminated samples to contaminated samples and aflatoxin contamination levels was modeled as a first order polynomial with an R 2 value of 0.9723. • Multispectral imaging detects the aflatoxin contamination in food. • Field deployable, economical, and scalable imaging setup for food quality control. • Machine learning models to classify different contamination levels accurately. • Bhattacharyya based regression model predicts contamination level at high resolution.
BibTeX
@article{madushan2024aflatoxin,
  author = {H.M.P.S. Madushan and Pasindu Malshan and K.K. Abewickrama and Vijitha Herath and Roshan Godaliyadda and M. P. B. Ekanayake and A. D. S. N. P. Athukorala and M.D.H.M. Kumari and Darsha Udayanga},
  title = {{Aflatoxin contamination level estimation in food using reflectance multispectral imaging based system}},
  journal = {Journal of Agriculture and Food Research},
  year = {2024},
  doi = {10.1016/j.jafr.2024.101401},
}

Loading related content...

Figure