Prediction of the quality properties and maturity of apricot by laser light backscattering imaging

The maturity level plays an essential role in the quality and shelf life of apricot. The present research aims to investigate the applicability and accuracy of the non-destructive laser light backscattering imaging method to predict the quality properties of apricot during ripening. The backscattering images of apricots were acquired at 650 nm in six stages of ripening. The images were segmented by two different thresholding techniques, and several space domain features were extracted from the segmented images. Artificial neural network (ANN), partial least squares regression (PLSR), and principle component analysis-artificial neural network (PCA-ANN) models were developed to predict the firmness and total soluble solids (TSS) of apricot using each of the extracted image features and their combination as input for the prediction models. Results revealed a high correlation between the extracted features from the backscattering images and the quality parameters of apricot during ripening. Modeling using ANN recorded better performance than PLSR. The highest coefficient of determination (R2) and the lowest root mean squared error (RMSE) of cross-validation were achieved with ANN as R2CV = 0.974, RMSECV = 3.482 and R2CV = 0.963, RMSECV = 1.146 for firmness and TSS, respectively. The results confirmed that the laser backscattering imaging method was successful in predicting the quality properties of apricot during ripening.