Monitoring and modeling of apple slices shrinkage during drying in hot air using texture features of laser backscattering images

The present paper investigated the feasibility of using texture-based features of laser backscattering images in monitoring and modeling of apple slices shrinkage during hot air drying. The backscattering imaging was performed at three wavelengths (650, 780 and 880 nm) in the visible and near-infrared regions. The acquired images were subjected to four texture analysis methods including first-order statistics of image histogram, co-occurrence matrix, gray level run-length matrix and wavelet transform. Stepwise multiple linear regressions was used to develop models and determine the most effective features by using individual types of feature sets and their combinations as inputs of calibration models. The results showed the capability of the texture features extracted from the laser backscattering images in the near-infrared wavelength range for prediction of apple slices shrinkage; by using homogeneity feature of co-occurrence matrix-90˚ at 880 nm (with Rp2=0.95, RMSEp=5.15) and fusion of the four feature sets extracted from different texture analysis methods at 780nm (with Rp2=0.94, RMSEp=5.61), could make models with high accuracy. This study showed that Laser backscattering imaging technique can be used as a non-destructive, rapid and low-cost method for prediction of the shrinkage in the process of hot air drying of apple slices.