Study On Quality Identification And Doping Analysis Of Olive Oil Based On Near Infrared Spectroscopy Technique

Despite the fact that olive oil market is prosperous in the world, many problems in oli ve oil market are increasingly serious such as non-standard labels, mixing the false with the genui ne. Many approaches are used for quality test ing, for inst ance, the physical and chemical test, ga s chromatogra phic method, l iquid chromatography.However, these approaches take a l ong process and their operations are compli cated.This thesi s combines near infrar ed spectroscopy and chemo metric resolut ion method to identify the quality of olive oil and to do quantitative analysi s of impurity content.The following research will be done:(1)Id ent ification method in qualitative analys is of olive oil based on near infrar ed spectroscopy is discussed in thi s thesis. First, wavelength intervals are selected from near infrared spectra data taking from Ex tra Virgin Olive Oil and Ordinary Virgin Olive Oil by synergy interval partial least square(si PLS); the features and perform ance models are evaluated by root mean s quare error of cross validation(RMSECV) and in the meanti me, opt imal wavelength intervals are s elected. Then,chara cter istic wavelengths are selecte d from them by it eratively retaining informative variables(IRIV), a principal com ponent analys is model is establis hed by using the chara cter istic wavelength of the selected wavelength. According to the researc h, 6chara cter istic wavelengths are selecte d by siPLS-IR IV-PCA and the total contribution rate of pri ncipal component one and two reac hed 98.65%. Less characteristic wavelengths are selected by siPLS-IRIV but t hey have highe r tot al contribution rate of principal component one and two compared with ot her m odels such as siPLS-PC A、siP LS-GA- P CA、siPLS-M C-UVE-PCA、s iPLS-CARS-PCA.(2)Quantitative analysis of impurity content in oli ve oil based on near infrared spectroscopy is studied in this thesis. Fi rst, pri ncipal component s of sample spectra are cal cula ted by PCA. Then, the principal components are used as the input of BP neural network to carry on the qualitative identificat ion. Furthe rly, optimal wavelength intervals are selected by siPLS algorithm and in the meantime,chara cter istic wavelengt hs are select ed from them by IRIV algori thm. PLSquant itative anal ysis model of olive oil adulteration i s established on the basis of the chara cter istic wavelengths. According to the research, 33 charact erist ic wavelengths are chosen by IRIV-siPLS model, the training set Rc = 0.9992, the RMSECV = 0.0056,the predi cti on set Rp = 0.9972, RMSEP = 0.0140. The IRIV-siPLS model has the highest stability and the best comprehe nsive performance compared with PLS、iPLS 、siPLS、 GA-P LS 、GA-siPLS and IRIV-PLS models.This research brings new ideas for the olive oil quality identificati on and doping analysis, it is of practical signifi cance t o improve product quality, st abil ity, and to standardize the market for olive oil.