Classification of internally damaged almond nuts using hyperspectral imagery

Hyperspectral transmission spectra of almond nuts are studied for discriminating internally damaged almond nuts from normal ones. We introduce a novel internally damaged almond detection method that requires only two sets of ratio features (the ratio of the responses at two different spectral bands) for classification. Our proposed method avoids exhaustively searching the whole feature space by first ordering the set of ratio features and then choosing the best ratio features based on the ordered set. Use of two sets of ratio features for classification is attractive, since it can be used in real-time practical multispectral sensor systems. Experimental results demonstrate that our method gives a higher classification rate than does use of the best feature selection subset of separate wavebands or than does use of feature extraction algorithms using all wavelength data.