The Study On The Application Of Multi-mode Vibrational Spectroscopy In The Detection Of Food Safety

Food safety represents one of the critical problems presented in Chinese publichealth and economy areas, attracting high concern of the whole society. Fooddetection technique is the key tool to solve the problem of food safety, and an accuratedetection technique will effectively prevent from the occurrence and development offood safety. Conventional detection methods of food safety is time-costing, expensive,require complex pre-processing, which is difficult to address the crisis of domesticfood issues nowadays, whereas the vibrational spectral technology is widely appliedin the field of detection for food safety due to its striking advantages such as beingfast, economical and user-friendly. However, the general detection is based on the solespectral technique whose limits might cause detection error and information leakage.Therefore, multi-mode vibrational spectroscopy is proposed, the theories of differentvibrational spectroscopies and those of relevant modeling are discussed, and wheatflour and dairy products are investigated. The main research in this thesis is asfollows:1. The strengths and weaknesses of conventional detection techniques andspectral ones were discussed in detail in the area of detection of food safety.According to this, near-infrared (NIR) spectroscopy and Raman spectroscopy wereselected as the target methods in the paper. Comprehensive analysis of the theory,strengths and weaknesses of multi-mode vibrational spectroscopy was included inorder to discuss the complimentary advantages in detection.2. Interference in the quantitative analysis of vibrational spectra of food sampleswas deeply discussed. Efforts were made to separate the background, the signal andnoise based on the difference between time and frequency components and thusmultiscale modeling were proposed in which sub-models were built according toinformation from different scales and given different weights, interference wassuppressed whereas useful information was made the best of and the integration ofpre-processing with modeling was accomplished.3. Borax in wheat flour was investigated through vibrational spectroscopy. Thetwo spectra were decomposed into multi scales and PLS models were built.Eventually high-level data fusion was used to make full use of the NIR and Raman spectra to quantitatively analyze the borax concentration of the samples in thevalidation set.4. Protein in milk powder was also investigated. NIR and Raman spectra of themilk powder samples were obtained and signal features of these spectra werediscussed to build the robust prediction model for protein detection.Multi vibrational spectroscopy was adopted and combined with decisive datafusion, the integrated diagnose by NIR and Raman spectroscopy was accomplished,which was successfully applied in the detection of the illegal additive of wheat flourand authentic protein in milk powder and is promising to be widely used.