Based on near infrared diffuse reflectance spectroscopy for the qualitative discrimination on food packaging materials polyethylene and polypropylene using different spectral pretreatment methods with different wavebands

principal component analysis(PCA) combined with three pattern recognition methods of SIMCA

Bayes discriminant and K-Nearest neighbor were adopted to establish three qualitative prediction models

and the prediction performances of the models were compared according to their correct recognition rates

in order to select the best model.Results showed that the qualitative correction models established by SIMCA

Bayes discriminant analysis and K-Nearest neighbor are better in the wavelength range of 1 050-1 550 nm.Six types of spectral preprocessing methods

i.e vector normalization

standard normal variable transformation

Centralization

moving average filtering

Savitzky-Golay filtering and first order differential combined with three pattern recognition methods of SIMCA

Bayes discrimination and K-Nearest neighbors were used to process the near infrared spectra of plastic samples.In the range of 1 050-1 550 nm

the principal component factor was 3.The qualitative correction model by K-Nearest neighbor using the original spectrum was the best

whose correct recognition rates for the sample's calibration set and prediction set were both 100%.It could provide a reference for the rapid identification of polyethylene and polypropylene as food packaging materials.