关键词: 近红外光谱, 转基因, 菜籽油, 小波去噪, 定性分析

Abstract: [Objective] The aim of this study was to establish the transgenic rapeseed oil identification models based on near infrared spectra, and study the effect of spectral preprocessing with wavelet denoising on the identification accuracy rate. [Method] Luhua, Jinlongyu and other six brands of bottled or barreled rapeseed oil altogether 117 samples have been already collected in earlier stage, which include 53 samples of transgene rapeseed oil and 64 samples of non-transgene rapeseed oil. The full spectrum of spectrum data of the 117 samples was collected by the FT-NIR analyzer of BRUKER company of Germany;the rapeseed oil spectral data was preprocessed by wavelet analysis, using db3 wavelet soft threshold to denoise the spectra;based on the near infrared spectral data of rapeseed oil samples, discriminant partial least squares(DPLS) were applied to set up the identification model of modified rapeseed oil. [Results] This study compared the accuracy of the modeling methods before and after the wavelet preprocessing. The accurate rate of the DPLS model increased from 96.43% to 100%. [Conclusion] The results indicated that wavelet denoising pretreatment can improve the accuracy of near infrared spectra of transgenic rapeseed oil identification model effectively.

Key words: near infrared spectroscopy, transgene, rapeseed oil, wavelet de-noising, qualitative analysis