Prediction models of texture attributes for radiofrequency assisted hot air drying of jujube slices based on morphological parameters

The effects of temperature and radiofrequency (RF) treatment on the texture attributes and morphological parameters of jujube slices were investigated using radiofrequency-assisted hot air drying (RFHAD). Drying temperature and RF treatment time have a significant influence on texture attributes such as chewiness and cohesiveness and morphological parameters such as area, roundness, and perimeter of jujube slices. The correlation between texture attributes and morphological parameters showed that area, roundness and perimeter were highly correlated with cohesiveness (P < 0.01), and chewiness was significantly correlated with circumference (P < 0.05). The regression model was estimated by using back propagation neural network (BPNN), partial least squares regression (PLSR), and multi-linear regression (MLR). Then, the texture attributes of jujube slices were predicted according to the changes of morphological parameters. BPNN predicted root mean square error (RMSE) of 1.1116 and 0.0091 for chewiness and cohesiveness, respectively, and coefficient of determination (R2) of 0.9943 and 0.9893 for chewiness and cohesiveness, respectively. In addition, BPNN obtained the highest R2 and lowest RMSE compared with PLSR and MLR. Therefore, BPNN can be used to explore the change mechanism of texture attributes of jujube slices from a macro point of view and provide a new idea for the drying technology of jujube slices.