Analysis of hyperspectral scattering image using wavelet transformation for assessing internal qualities of apple fruit

Hyperspectral scattering technique can provide an effective means for nondestructive measurement of fruit internal quality, while hyperspectral scattering image contains a lot of data which need effective data reduction. This research investigated 600 images of `Golden Delicious´ apple and decomposed 101 wavelengths into 2 layer and 3 layer using Db3 of Danbechies wavelet series as basis function. The low frequency wavelet coefficients were selected as input coupled with multiple linear regression (MLR) algorithm and partial least squares (PLS) algorithm to develop the prediction model of apple internal qualities. The simulation results show that both the accuracy and standard error of prediction model developed by features extract from 2 layer wavelet transformation are better than that by traditional way of feature waveband selection, no matter fruit firmness or soluble solids content.