微型近红外光谱法快速检测赛买提杏糖度  

Rapid detection of Cymaiti apricot soluble solids content by miniature near infrared spectroscopy

作　　者：彭娟 邵学广[2] 楚刚辉 PENG Juan;SHAO Xueguang;CHU Ganghui(Chemistry Laboratory of Xinjiang Special Medicinal Edible Plant Resources,College of Chemistry and Environmental Science,Kashi University,Kashi 844000,China;Analytical Science Research Center,School of Chemistry,Nankai University,Tianjin 300071,China)

机构地区：[1]新疆特色药食用植物资源化学实验室,喀什大学化学与环境科学学院,喀什844000 [2]南开大学化学学院分析科学研究中心,天津300071

出　　处：《分析试验室》2023年第10期1285-1291,共7页Chinese Journal of Analysis Laboratory

基　　金：国家自然基金项目(22174075);新疆维吾尔自治区高校科研计划自然科学重点资助项目(XJEDU2020I016)资助。

摘　　要：为实现赛买提杏糖度(SSC)的快速检测,并建立一个稳健的预测模型,本文使用微型近红外光谱仪采集光谱,数显糖度计测定SSC,对比了连续小波变换(CWT)、卷积平滑(SG)、多元散射校正(MSC)、标准正态变换(SNV)、一阶导(1st),二阶导(2nd)及其两两组合的光谱预处理方法建模结果,结果显示,CWT,CWT+MSC和CWT+SNV 3种方法结果较好。以这3种预处理方法为基础,对比了竞争自适应加权抽样(CARS)、随机化试验(RT)、蒙特卡罗-无信息变量消除法(MC-UVE)和C值法4种变量选择方法,分别得出与杏糖度相关的光谱变量,建立了赛买提杏糖度偏最小二乘(PLS)预测模型,经CWT+MSC光谱预处理加C值法变量选择建立的PLS模型最佳,其预测相关系数(Rp)从0.844优化到0.960,预测均方根误差(RMSEP)从1.273降低到0.699,相对分析偏差(RPD)从1.853上升到3.430。结果表明,适当光谱预处理和变量选择能有效提高赛买提杏糖度快速无损检测的效率和准确性。To realize the rapid detection of Cymaiti apricot soluble solids content(SSC) and establish a steady prediction model of SSC,a miniature near-infrared spectrometer was used to collect the spectra,and a digital display sugar meter was used to determine the SSC.Then the modeling results including continuous wavelet transform(CWT),Savitzky-Golay(SG smoothing),multiplicative scatter correction(MSC),standard normal variate(SNV),firstderivative(1st),second derivative(2nd) and the spectral pretreatment method of two combinations were compared,and the models of CWT,CWT+MSC and CWT+SNV showed good results.Based on above three pretreatment methods,four variable selection methods including competitive adaptive reweighted sampling(CARS),randomization test(RT),informative variables elimination(MC-UVE) and C value were compared by obtaining the spectral variables related to SSC of apricot.And the partial least squares(PLS) prediction model of apricot SSC was established.The best model was established by CWT+MSC peretreatment combined with C value variable selection method.The correlation coefficient(Rp) of prediction set increased from 0.844 to 0.960,the root mean square error of prediction(RMSEP) set decreased from 1.273 to 0.699,and the relative percent deviation(RPD) increased from 1.853 to 3.430.The results showed that spectral pretreatement and variable selection properly could effectively improve the efficiency and accuracy of the portable and rapid nondestructive testing of apricot SSC.

关 键 词：赛买提杏 变量选择 近红外光谱 糖度 

分 类 号：O657.3[理学—分析化学]