食品科学

• 安全检测 • 上一篇    下一篇

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,葛慧芳,于殿宇   

  1. 1.哈尔滨商业大学计算机与信息工程学院,黑龙江 哈尔滨 150028;2.东北农业大学食品学院,黑龙江 哈尔滨 150030
  • 出版日期:2016-08-25 发布日期:2016-08-30
  • 通讯作者: 于殿宇
  • 基金资助:

    国家自然科学基金面上项目(31271886)

Optimization of Characteristic Wavelength Variables of Near Infrared Spectroscopy for Detecting Edible Oil Acid Value

WANG Liqi, LIU Yanan, ZHANG Qing, CUI Yue, GE Huifang, YU Dianyu   

  1. 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
  • Online:2016-08-25 Published:2016-08-30
  • Contact: YU Dianyu

摘要:

以近红外光谱快速检测大豆油脂酸值为目标,研究间隔偏最小二乘(interval partial least square,iPLS)结合遗传算法(genetic algorithm,GA)及连续投影算法(successive 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4 000~12 000 cm-1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4 540~5 346 cm-1和6 807~7 004 cm-1组合特征波段,R2和预测均方根误差(rootmean square error of prediction,RMSEP)分别为0.978 9和0.064 3;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SPA 2 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PLS校正模型,其R2和RMSEP分别为0.985 9和0.045 1。研究表明,在油脂酸值近红外光谱分析中,采用iPLS-GASPA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,可为快速无损检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

Abstract:

With the goal of achieving rapid detection of soybean oil acid value by using near-infrared (NIR) spectroscopy,
this study optimized the selection of the characteristic wavelength variables by combined use of interval partial least square
(iPLS), genetic algorithm (GA) and successive projection algorithm (SPA). A total of 100 soybean oil samples with different
acid values were collected, their NIR transmittance spectra in the range of 4 000–12 000 cm-1 were acquired. Firstly, the
characteristic wavebands of 4 540–5 346 cm-1and 6 807–7 004 cm-1 were extracted from the original spectra by iPLS, with a
determination coefficient (R2) and root mean square error of prediction (RMSEP) of 0.978 9 and 0.064 3, respectively. Then,
the characteristic wavelength variables closely related to oil acid value were selected by GA and SPA from the previously
selected bands, respectively. It was shown that the PLS calibration model established using 12 variables consisting of the top
6 characteristic wavelengths from optimal selection results of each of the two algorithms was optimum, with R2 and RMSEP
of 0.985 9 and 0.045 1, respectively. The research indicated that selection of the characteristic wavelength variables by
iPLS-GA-SPA in NIR analysis for oil acid value could effectively remove redundant information, and decrease the
complexity of the model. This paper can offer important reference for rapid and non-destructive detection of oil acid value.

Key words: oil acid value, near infrared spectroscopy (NIR), characteristic wavelength variables, interval partial least square (iPLS), genetic algorithm (GA), successive projection algorithm (SPA)

中图分类号: