Pears characteristics (soluble solids content and firmness prediction, varieties) testing methods based on visible-near infrared hyperspectral imaging

Visible-near infrared hyperspectral images (400–1000nm) were used for non-destructive variety discrimination and prediction of soluble solids content (SSC) and firmness of pears. An imaging spectroscopy system was assembled to acquire scattering images from pears. Spectra of 180 pear samples from three varieties were analyzed by four algorithms of principal component analysis (PCA), partial least squares (PLS), successive projections algorithm (SPA) and Fisher linear discriminant analysis (Fisher LDA) to detect SSC, firmness and varieties of pears. Then PLS models under whole spectral wavelengths were compared with SPA-PLS models under effective wavelengths. The SPA-PLS models were considered to be the best method for detecting firmness and SSC of pears. The model led to correlation coefficient (r) of 0.9977 for firmness and 0.9924 for SSC and root mean square error (RMSEP) estimated by cross-validation of 0.062653 for firmness and 0.03175 for SSC. The correct answer rate of 95.56% for variety discrimination was achieved by Fisher LDA.