Research On Pattern Recognition And Blended Proportion Detection Of Vegetable Oil

As one of main nutriments in people's daily life, edible vegetable oils have many kinds and categories. Their components, physical and chemical characteristics are diversified by effect of different raw species, origin, production process and other factors. So the detection of vegetable oil is difficult. Conventional physical and chemical detection methods are difficult to meet the requirements of contemporary laws and regulations,well, modern apparatus detection methods exist some other problems, such as more complex operations, the high cost of testing, the less universal in our country. Therefore, it is very important to research and develope a simple, accurate and rapid detection method of vegetable oils, to achieve qualitative and quantitative detection.In this paper, peanut oil, maize oil, soybean oil, sunflower oil, palm oil and their binary blended oils were taken as main research objects. Relyed on the ultraviolet spectrum of vegetable oils, cluster analysis, principal component analysis, discriminant analysis were used to study the recognition method of single vegetable oils; template matching method was applicated to recognise the binary blended vegetable oils; the blended proportion were analyzed through linear regression, multiple linear regression, principal component regression and partial least-squares regression. The recognition program of single and binary blended vegetable oils and the detection program of blended proportion were established in Matlab platform; the identification and detection interface of vegetable oils was founded; the identification and detection of vegetable oils was achieved rapidly.The main contents and specific conclusions of this paper as follows:1) Theoretical analysis of the UV spectra of the nature of vegetable oils was discussed, conjugated diene acids and polyunsaturated acids contained in vegetable oils have peaks detection between 220nm and 300nm, the vitamin A at 325nm, vitamin D at 265nm, vitamin E at 285nm have peaks detected, sterols of vegetable oils absorbe ultraviolet light too. The interaction and mutual influence of these components form the UV absorption spectrum of vegetable oils.2) The UV absorption spectrum characteristics of vegetable oils were gotten through scan of UV absorption spectrum curve. Six vegetable oils had peaks detection at 271, 267, 258, 252nm, acromion was detected at 245nm. Sesame oil's peak detected at 282nm,while other oils'at 279nm. peanut oil, soybean oil, sunflower oil and palm oil had peak detection at 227nm,while corn oil and sesame oil at 230nm.3) The first derivative and second derivative processing were used to analyze the UV absorption spectra of six kinds of vegetable oils. The first derivative spectrometry of sesame oil had characteristic trough detection at 297nm, second derivative spectrometry had characteristic peak detection at 299nm and characteristic trough at 294nm, while other five kinds of vegetable oils had no peak or trough detection at these wavelengths. The first derivative spectra of six kinds of vegetable oils had peaks detection at 221,264nm and troughs detection at 242, 248, 284nm; The second derivative spectra had peaks detection at 219, 249, 262nm and troughs detection at 223, 246, 266nm.4) Comparison of the ultraviolet absorption spectrum and derivative spectrum of vegetable oils, except characteristic peaks and troughs of sesame oil, the other peaks and troughs of six kinds of vegetable oils appeared in the same or similar position, however, absorbance values were different, which show that oils contain the same absorption material, well difference in content. At the same concentration, Corn oil has the strongest absorbance between 220 and 250nm; between 250 and 300nm, the absorption spectra of corn oil and sunflower oil was close to each other. In the entire scan range, the absorbance of peanut oil was the smallest.5) Under the concentration of 2 mg/mL, the UV absorption spectrum of solution could reflect the absorption characteristics of six vegetable oils, all the peaks and troughs of spectrometry were detected. The absorbance value was appropriate and fluctuation scope of absorbance was small, the deviation was less than 5%, so, 2 mg/mL was selected as the best experimental concentration.6) Based on the original absorbance spectra, first derivative absorbance and second derivative absorbance, cluster analysis was applied to analyze six kinds of vegetable oils. Comparison result, cluster analysis of the first derivative absorbance was the best, corn oil and sesame oil were clustered respectively, soybean oil and sunflower oil were clustered together for a class, peanut oil and palm oil for a class.7) The absorbance under wavelength of 290, 279, 271, 267, 258, 252, 245, 227nm was analysed through the method of principal component analysis, three main components was got and their cumulative contribution rate was 97.31%, containing most information of sample. The first principal component took up 76.47%, the second 12.7%, the third 8.14%. A271, A267, A252, A258, A245 and A279 had the significant contribution to the first principal component, A290 contributed more to the second principal component, A227 had significant contribution to the third principal component. Three principal component score function were founded. According to principal component scores, the distribution of the vegetable oil samples was gotten,According to the regional distribution of scores points, the type of vegetable oil. could be jundged.8) The absorbance of six kinds of vegetable oils was analysed by means of progressive discrimination analyses. A290,A279,A252,A245,A227 were selected as the main variable and got three kinds of discriminant function. Through comparison, the best discriminant function was linear discriminant function.. Substituted the absorbance into the discriminant function to calculate the function value, the maximum determined the category of vegetable oil. Verification by experiment, the accuracy of discriminant function was 100%.9) This paper set sesame peanut blended oil as example. Through the establishment of one dimension regression equation, multiple regression equation, principal component regression equation and partial least-squares regression equation, analysed the best way to calculate the mixing amount of peanut oil.The best regression equation, identified by comparison, was one dimension regression equation, which was based on the original UV-absorbance. The equation was y=-2.8706A290+1.1319. Validation results were satisfactory and the detection limit for peanut oil was 8%, the detection range was 8% -92%, the maximum detection error was 5.4%.10) According to the principles of template matching method, template file of the UV absorbance of six single oils was founded in Matlab platform, contained UV absorbance of six kinds of vegetable oil. Pattern recognition program of vegetable oils was established, the recognition of vegetable oil was realized and the accuracy was 100%.11) The structure template, contained the name of binary mixtures of oil, UV absorb- ance and coefficient of the regression equation, was founded in Matlab platform. Pattern recognition program was complied and the blended proportion was calculated in program, achieved qualitative identification and quantitative analysis of binary blended oils, the accuracy qualitative identification was 92.3%.12) The recognition interface of vegetable oils was founded through GUIDE editor of Matlab. The interface structure was smart and operated conveniently. The identification of vegetable oils and the detection of blended proportion were achieved rapidly.