Industrial quality inspection using image analysis on astaxanthin coating in aquaculture feed pellets is of great importance for automatic production control. The pellets were divided into two groups: one with pellets coated using synthetic astaxanthin in fish oil and the other with pellets coated only with fish oil. In this study, multispectral image analysis of pellets captured reflection in 20 wavelengths (385–1050 nm). Linear discriminant analysis (LDA), principal component analysis, and support vector machine were used as statistical analysis. The features extracted from the multispectral images were pixel spectral values as well as using summary statistics such as the mean or median value of each pellet. Classification using LDA on pellet mean or median values showed overall good results. Multispectral imaging is a promising technique for noninvasive on-line quality food and feed products with optimal use of pigment and minimum amount of waste.
Martin Georg Ljungqvist, Bjarne Kjær Ersbøll, Michael Engelbrecht Nielsen, Stina Frosch, "Multispectral Image Analysis for Astaxanthin Coating Classification" in Journal of Imaging Science and Technology, 2012, pp 20403-1 - 20403-6, https://doi.org/10.2352/J.ImagingSci.Technol.2012.56.2.020403