Laser-induced fluorescence spectroscopy enhancing pistachio nut quality screening

Food safety and quality has become increasingly important in our society, driving the development of novel optical food sensing technologies. Optical sensing technologies have shown good product identification performances, while offering a non-destructive measurement and the possibility for in-line applications. They are, however, currently often limited in their sensitivity and product variability. We therefore pursue the development of a novel pistachio nut screening methodology offering a multi-defect detection, simultaneously detecting shells, tree parts and aflatoxins, by combining fluorescence spectroscopy with advanced chemometrics. Specifically, both one- and two-photon induced fluorescence are investigated, in combination with Linear Discriminant Analysis, Quadratic Discriminant Analysis and K-Nearest Neighbors algorithm, enabling to optimize both the hardware and software parameters. Optimal results were obtained combining the fluorescence spectra using 385 nm excitation with Quadratic Discriminant Analysis, showing a classification accuracy of 99.2% for the healthy pistachio kernels, together with a false positive rate of only 0.8%. This excellent classification accuracy, while considering a multi-defect challenge, is exceeding the state-of-the-art, paving the way towards an improved pistachio screening, benefitting the food processing industry.