Hyperspectral imaging system to online monitoring the soy flour content in a functional pasta

Pasta enriched with soy flour can be considered as a functional food, due to its content in nutraceutical compounds such as isoflavones, carotenoids, and other antioxidants. The quantification of the amount of a functional ingredient is an important step in food authenticity. The availability of non-destructive techniques for quantitative and qualitative analyses of food is therefore desirable. This research aimed to investigate the feasibility of hyper-spectral imaging in reflectance mode for the evaluation of the soy flour content, also to investigate the possibility of implementing a feed-back control system to precisely dose the soy flour during the industrial production of pasta. Samples of pasta in shape of spaghetti were produced with durum wheat semolina and soy flour at increasing percentages (0, to 50%, steps of 5%). A feature selection algorithm was used to predict the amount of soy flour. The most influent wavelengths were selected, and a six-term Gauss function was trained, validated, and tested. The identified transfer function was able to predict the percentage of soy flour with high accuracy, with an R-adj(2) value of 0.98 and a Root Mean Square Error of 1.31. The developed system could represent a feasible tool to control the process in a continuous mode.