Abstract: Near infrared spectroscopy combined with chemometrics was used to trace geographical origin of peanuts from different provinces of China. The original spectra were subjected to noise reduction by standard normal variate (SNV) and de-trend (DT) transformation (SNV+DT), and dimension reduction by principal component analysis; wavelet transform (WT) was also used for noise reduction and data compression. The extracted features were processed with linear discriminant analysis, Bayes discriminant analysis and k-nearest neighbors to discriminate the geographical origin of peanuts produced in different provinces of China. The results showed that WT and k-nearest neighbors present the best classification. Approximately 100.0% samples were correctly classified in original test, and 55.9% samples were correctly classified in cross-validation test. As a result, the geographical origin of peanuts from different provinces of China was discriminated. However, the performance of the discrimination models still needs to be improved.

Key words: peanut, near infrared spectroscopy, geographical origin, principal component analysis, wavelet transform, discriminant analysis