NIR Calibration Model Building Of Oil Content And Components Of Symplocos Paniculata Fruit

Symplocos paniculata, a member of Symplocaceae, is deciduous stunned or shrub, which is an very important ecological and economic resource and has great potential development for its high oil content in the fruit, among which oleic acid and linoleic acid are the major component but constitutes of low stearic acid. However, relatively little work has been done on the fast and efficient detection methods for Symplocos paniculata in the aspect of selective improved varieties and large-scale processing use. Therefore, Symplocos paniculatain fruits collected from different places were used as experimental materials, and the best set of the instruments and the choices of the pretreatment and the model regression method were studied by means of oil conventional detection methods, gas chromatography-mass spectrometry analysis,near infrared spectrum detection technology and other methods to bulid a rapid detection method for the inclusion and lay the foundation for the rapid detection of oil content and fruit quality and speeding upexcellent resources breeding. The results were showed as follows:(1)Compared with the different oil models which were bulit with the distinguishability were 2,5,10nm and the times of reinstaling samples were1,2,3 and scanning 20 times in each install, we found when the scanning parameters were 5nm, 3 times reinstall, the model is best.(2)Eliminating abnormal samples could further optimize the quantitative calibration model. The correlation coefficient was increased to 0.9733, the interaction validation decision coefficient was up to 0.9474, rather the RMSECV was reduced to 0.0427 in the same time. In addition,5 kinds of fatty acids quantitative calibration models were remarkably improved as well. What mentioned above could be found after eliminating the abnormal points of leverag as well as residual(3)The different spectrum pretreatment method could have effect on model construction, therefore,the stability and ability of model could be greatly improved because of the appropriate spectrum pretreatment method.The results showed that the SNV+2st Der works best to the oil cotent model. The corresponding optimal pretreatment combinations to palmitic acid, stearic acid, oleic acid, linoleic acid, linolenic acid respectively were SG smoothing(11,7)+1st Der, MSC+2st Der, SG smoothing(11,7)+lst Der, MSC+2st Der and 1st Der,respectively.(4)The modeling methods of PLS was better than PCR by comparing with each other. The results presented that both internal validation decision coefficient and forecast decision coefficient of oil content, palmitic acid, oleic acid and linoleic acid were all above 0.9, and the RMSECV and RMSEP were small at the same time, which could be used at pratical production. Although the determination coefficients were more than 0.8, the modes of stearic acid and linolenic acid were not stable enough in the predictive ability.