Development of a general model for monitoring moisture distribution of four vegetables undergoing microwave-vacuum drying by hyperspectral imaging

A hyperspectral imaging system was employed to develop a general model for monitoring moisture distribution of four types of vegetables undergoing microwave-vacuum drying (MVD). Partial least squares regression and least squares support vector machines were developed based on full and optimal wavelengths, while the multiple linear regression (MLR) was built from optimal spectral data. Genetic algorithm-MLR (GA-MLR) model with a determination coefficient in prediction of 0.974 and a root mean square error in prediction of 4.70% was used for the moisture content visualization of four MVD vegetables. Moisture distribution maps indicated a similar center drying pattern among different vegetables, while the conner effects occurred in celery stalks and spinach leaves. Additionally, evaluations of color changes of the vegetables using the computer vision system indicated a decrease of the browning index in dried carrot slices and celery stalks but an increase in potato slices and spinach leaves.