Vis-NIR Spectroscopy for Non-destructive Classification of Juicy Peach

A new method for discrimination of juicy peach varieties by means of Vis-NIR spectroscopy as a non-destructive measurement technique was developed. A relation was established between the reflectance spectra and varieties. Different pretreatment techniques were compared. 3 segments of moving average smoothing and first derivative (average number of left-rightside was 1) is the best. PCA provided the loading plot of PC1 and more optimal waveband between 400 to 750 nm was obtained. PCA compressed spectral data into a small quantity of PCs. The scores of the first 16 PCs had been applied as inputs to stepwise discriminant analysis (SDA). 75 samples of three varieties were selected random and used to build PCA-SDA model. This model was used to predict the varieties of 15 unknown samples and repeated 3 times. The recognition rate is 93.3%, better than by only using the PCA model. So PCA-SDA model is practicable, precise and reliable