食用油脂酸值近红外光谱特征波长优选.pdf

※安金检
2016,Vb37No.16205
食用油脂酸值近红外光谱特征波长优选
王立琦,刘亚楠,张青,崔月',葛煮芳',于殿宇
(1.哈尔滨商业大学计算机与信息工程学院,黑龙江哈尔滨150082.东北农业大学食品学院,黑龙江哈尔滨1500
摘要:以近红外光谱快速检测大豆油脂酸值为目标,研究间隔悩最小二乘( interval partial least square,iPLS)结
合遗传算法( genetic algorithm,GA)及连续投影算法( succesive projection algorithm,SPA)的特征波长变量优选
方法。制备不同酸值的大豆油脂样品100个,并在4000~12000cm'范围内采集了油样的近红外透射光谱。首先用
iPLS法从原始光谱中初步筛选出4540~5346m1和6807~7004cm~'组合特征波段,R2和预测均方根误差(rot
square error of prediction, RMSEPY分别为09789和0.0643;然后分别用GA和SPA从特征光谱区域中筛选出
与油脂酸值密切相关的特征波长变量,从GA和SPA2种选择结果中各选取前6个波长点,以12个特征波长变量建
立PIS校正模型,其和 RMSEP4分别为0.9859和0.0451。研究表明,在油脂酸值近红外光谱分析中,采用 IPLS-GA-
SPA相结合的方法进行特征波长选择能有效去除冗余信息,降低模型复杂度,可为快速无损裣测油脂酸值提供重要
理论依据。
关键词:油脂酸值;近红外光谱:特征波长变量:间隔偏最小二乘:遗传算法;连续投影算法
miz on of a ae n h a l of Near I na y for Deecting E ble il Acid al
WANG Ligi, Lru Yanan, ZHANG Qing CUI Yue, GE Huifang YU Dianyu
(1. School of Computer and Information Engineering, Harbin University of Commerce, Harbin 150028, China:
2. School of Food Science and Technology, Northeast Agricultural University, Harbin 150030, China
Abstract: ith the goal of achieving rapid eection of ybean oil a value by using ea-infrardnrspecroscopy
this study optimized the selection of the characteristic wavelength variables by combined useof interval patia least qua
(P e l TMG and c ssive p e n a PA A al of 100 oy ean i mples with t
acid values were cold eir NR a m a a e an of cm we. s, the
carac esc w ands of 4 434 a 8_ 4cm w x me l a y PLS, with a
determinaion o fc a m of RMP of7a .43 vely The
the characteristic wallen h abes closely elad oil acid value were selected by GA and SPA from the previously
selected bands, respectivel It was hown that the PLS calibration model established using I2 variablesconsisting of the top
6 characterisic wavelengths from ptimal s lectin resuls of each of the two algorithms was optimum, with R and RMSEP
of 0.9859 and 0.045 1, respectively. The research ndia ed that selection of the characteristic wavelength variables b!
IPLS-GA-SPA in NIR analysis for oil acid value could effectively remove redundant information, and decrease th
complexity of the model. This paper can of er imp ant r erence for apid a structive detection of oil acid value
eywords:oilacidyalue;nearinfraredspectroscpy(NIR);characteristieavelengthvariablessintervalpartialastsquare
(IPLS) genetic algorithm(GA) successive projection algorithm(SPA
DOR10.7506spkx1002-663020161603
中图分类号:TS
文献标志码;A
文章编号:1002-6630(2016)16-0205-06
引文格式
王立琦,刘亚楠,张青,等食用油脂酸值近红外光谱特征波长优选叮.食品科学,2016,37(16):205-210.DO:10.7506
spkx1002-6630-201616033.http://www.spk.net.n
WANG Lig, LIU Yanan, ZHANG Qing, et all. Optimization of characteristic wavelength variables of near infrared
spectroscopy for detecting e e oil acid value Food Science 2016, 37(16 05-210(n Chinese with English abstract
Do:10.750/spx1002-6630-201616033.http://www.spkax.net.cn
收稿日期:2016-03-16
基金项目;国家自然科学基金面上项目(31271886)
作者简介:王立琦(1966-),女,教授,博士,主要从事食品安全快速检测研宄。E-mail;hsdl163.com
通信作者:于殿宇(1964ー-),男,教授,博士,主要从事粮油加工技术研究。E-mail:dyu2000126com
万方数据