运用近红外光谱技术快速测定植物食用油的过氧化值  

Rapid determination of peroxide value of plant by near infrared spectroscopy

作　　者：彭博[1] 涂斌[1] 陈志[1] 郑晓[1] 

机构地区：[1]武汉轻工大学机械工程学院,湖北武汉430023

出　　处：《武汉轻工大学学报》2016年第1期13-16,42,共5页Journal of Wuhan Polytechnic University

摘　　要：采用近红外光谱技术建立了一种快速准确的植物食用油过氧化值预测方法。收集6种共61份植物食用油样品采集其近红外光谱数据,在Matlab平台上对光谱数据进行处理建模。采用多元散射校正(MSC)、正交信号校正(OSC)、标准正态变量变化和去趋势技术联用算法(SNV-DT)分别对原始光谱进行预处理,并使用支持向量机(SVM,Support Vector Machine)建立回归模型,其中分别采用网格搜索算法(CV)以及粒子群算法(PSO)对SVM惩罚参数C和RBF核函数参数g进行优化建模,旨在找出一种光谱预处理和参数优化联用的方法。实验结果表明在SNV-DT与粒子群算法联用的这种方法要优于其他,预测集和校正集的相关系数分别为90.139%和99.999%,并且成功的从61份植物食用油样品中分辨出了14份过氧化值超标油。实验结果证明了采用近红外光谱技术可以快速准确的预测植物食用油的过氧化值,实现对植物食用油合格与否进行判别,具有较强的实用价值和推广价值。Using near infrared spectroscopy technology to build a fast and accurate method for predicting the peroxide value of plant oil. We collected a total of 61 six kinds of plant oil samples collected its near-infrared spectral data and modeling the spectral data in the Matlab platform. The original spectra were pretreated by using multiple scatter correction（ MSC）,orthogonal signal correction（ OSC）,standard normal variable variation and De-trending（ SNV-DT）. And using support vector machine（ SVM,Support Vector Machine） to build regression model. The SVM penalty parameters C and RBF kernel function parameters G are optimized by using the grid search algorithm（ CV） and particle swarm optimization（ PSO）. It aims to find a spectral preprocessing and parameter optimization in combination with the methods. Experimental results show that this method SNV-DT and particle swarm algorithm combined with better than others. Prediction set and calibration set correlation coefficients were 90. 139% and 99.999%,and the successful resolution of edible oils from 61 plant samples out of the 14 parts of the peroxide value exceeded oil. Experimental results show that using near infrared spectroscopy can quickly and accurately predict the peroxide value of plant oil,oil plants to achieve the qualified or not to discriminate,with a strong practical valueand promotional value. Experimental results prove the using near infrared spectroscopy technology can quickly and accurately predict the peroxide value,realization of plant oil is qualified or not to judge,has a strong practical value and popularization value.

关 键 词：植物油 近红外光谱技术 支持向量机 过氧化值 

分 类 号：TS227[轻工技术与工程—粮食、油脂及植物蛋白工程]