Non-destructive detection of bread staleness using hyperspectral images

Hyperspectral imaging, a technology that combines imaging and spectroscopy, provides extensive spatial and spectral information, simultaneously. It is currently being developed as a non-destructive and rapid diagnostic tool for assessing food quality and safety. In this study, hyperspectral imaging was utilized to investigate the process of bread staleness and its effect on the behavior of the bread crumb within the wavelength range of 950-400 nm and with a resolution of 0.795 nm. Principal components were extracted and three modeling methods - PCR, PLSR, and GRNN - were employed to predict texture characteristics during six days of storage. Based on the findings of this study, it was observed that the General Regression Neural Network (GRNN) method demonstrated superior performance in terms of R2 values for both springiness and stiffness, with values of 0.96 and 0.94, respectively. Furthermore, the GRNN method also exhibited the lowest Root Mean Square Error (RMSE) values for cohesiveness and stiffness, with values of 0.11 and 0.32, respectively. This demonstrates the capability of the generalized regression neural network model to predict the textural characteristics of bread.