Soluble solids content and pH prediction and varieties discrimination of grapes based on visible–near infrared spectroscopy

Visible and near infrared reflectance spectroscopy was used for the non-destructive variety discrimination and the prediction of soluble solids content (SSC) and pH of grapes. Spectra of 439 grape samples from three varieties were analyzed by genetic algorithm (GA). Results of GA indicated the most effective wavelengths for the variety discrimination and content prediction. With the spectra of the effective wavelengths from GA, variety discrimination and component content prediction were executed based on least-squares support vector machine. There were 293 samples of grapes used for calibration and the left 146 samples for prediction. The correct answer rate of 96.58% for variety discrimination was achieved. The correlation coefficients are 0.9781 for pH and 0.9065 for SSC in the prediction set. Models established based on effective wavelengths overperformed those that use whole spectral data.