APPLICATION OF NEAR INFRARED HYPERSPECTRAL IMAGING FOR ANALYSIS OF TEXTURIZED VEGETABLE PROTEINS

In recent years, an increasing interest in vegetable proteins has been observed both from the consumers side and from the industrial one. This is demonstrated by the conspicuous investments for the plant-based sector which reached 3.1 billion dollars in 2020. Texturized vegetable proteins (TVP) are quantitatively the most important ingredient for the production of plant-based products. TVP imitate the fibrillar structure of meat muscle and strongly affect the nutritional and textural properties of the final product. Considering the increasing usage of TVP for a wide spectrum of applications in the industrial context, a critical aspect regards the development of suitable and convenient methods for quality control. Proximate composition is the first mandatory aspect to be monitored but is a demanding task which requires time, polluting solvents, and resources. By contrast, near infrared spectroscopy (NIRS) has been proven to be an efficient technique for rapid and non-destructive analysis. Furthermore, when NIRS is coupled with imaging techniques, a comprehensive spatial and spectral information of the product under study could be achieved. In this regard, Hyperspectral Imaging (HIS) has been recently suggested as a promising non-destructive technique for evaluating the quality of TVP and plant-based products. In this framework, this work is aimed at studying the feasibility of NIR-HSI for the analysis of TVP chemical composition. Four different TVP have been produced in duplicate by a low-moisture extrusion process, combining different protein sources and analysed for total protein content, total fat and ashes. NIR hyperspectral images were collected in reflectance mode by using a spectrometer (Headwall photonics model 1002A-00371) working in the wavelength range of 1009-1694 nm with a spectral resolution of 4.85 nm and spatial resolution of 30 μm. After acquisition, the spectral images were processed in Matlab environment by using the PLS_toolbox and HYPER-Tool. Data were explored using PCA and then subjected to regression analysis by PLS1 algorithm. The figures of merit of the regressions in calibration and cross-validation showed excellent performance of the developed models, with values of R2 always higher than 0.90 and low values of RMSE. Prediction of external test set confirms these results.