Food quality assessment and quantification using multispectral images from screen and smartphone for kiwifruit ripeness and rice discrimination

High-quality food products require precise, rapid, and objective determination of their quality. Therefore, spectral imaging stands out as one of the most powerful and fastest non-destructive tools for the analysis and quality control of food materials. In this study, two scenarios were conducted. The first scenario involved assessing the ripeness of kiwi fruit, while the second scenario focused on the classification of various types of Iranian rice using a smartphone setup, along with chemometric methods including Partial Least Squares (PLS) and Partial Least Squares Discriminant Analysis (PLS-DA). In this study, a simple setup based on a smartphone and laptop monitor was designed to enable the capture of multispectral images. The performance of PCR and PLS models was evaluated based on their RMSEC and RMSEP values. The PCR model showed RMSEC of 3.2336 and RMSEP of 4.9210, indicating a high level of accuracy in predicting the time elapsed since cutting. On the other hand, the PLS model demonstrated superior performance with RMSEC of 0.9244 and RMSEP of 0.6073. Additionally, the coefficient of determination (R²C = 0.94 and R² P = 0.88) for the PLS models was calculated to assess the goodness of fit, and RMSECV values were calculated for the classification of group pairs using the leave-one-out cross-validation technique. The calculated RMSECV values were 0.0253 for the pairs of Anbarbu-Domsiyah groups, 0.0274 for Anbarbu-Gohar groups, 0.0222 for Tarom-Gohar groups, and 0.0200 for Domsiyah-Gohar groups. The four different types of Iranian rice were successfully discriminated.