• EI
  • Scopus
  • 食品科学与工程领域高质量科技期刊分级目录第一方阵T1
  • DOAJ
  • EBSCO
  • 北大核心期刊
  • 中国核心学术期刊RCCSE
  • JST China
  • FSTA
  • 中国精品科技期刊
  • 中国农业核心期刊
  • CA
  • WJCI
  • 中国科技核心期刊CSTPCD
  • 中国生物医学SinoMed
中国精品科技期刊2020
张丽华, 郝莉花, 李顺峰, 纵伟. 基于支持向量机的近红外光谱技术快速鉴别掺假羊肉[J]. 食品工业科技, 2015, (23): 289-293. DOI: 10.13386/j.issn1002-0306.2015.23.051
引用本文: 张丽华, 郝莉花, 李顺峰, 纵伟. 基于支持向量机的近红外光谱技术快速鉴别掺假羊肉[J]. 食品工业科技, 2015, (23): 289-293. DOI: 10.13386/j.issn1002-0306.2015.23.051
Zhang Li-hua, Hao Li-hua, Li Shun-feng, Zong Wei. Fast discriminating the adulteration of minced mutton with near infrared spectroscopy based on support vector machine[J]. Science and Technology of Food Industry, 2015, (23): 289-293. DOI: 10.13386/j.issn1002-0306.2015.23.051
Citation: Zhang Li-hua, Hao Li-hua, Li Shun-feng, Zong Wei. Fast discriminating the adulteration of minced mutton with near infrared spectroscopy based on support vector machine[J]. Science and Technology of Food Industry, 2015, (23): 289-293. DOI: 10.13386/j.issn1002-0306.2015.23.051

基于支持向量机的近红外光谱技术快速鉴别掺假羊肉

Fast discriminating the adulteration of minced mutton with near infrared spectroscopy based on support vector machine

  • 摘要: 为实现掺假羊肉的无损鉴别,利用傅里叶变换近红外光谱分析技术建立混入鸭肉的掺假羊肉糜的快速检测方法。实验通过在羊肉糜中添加不同比例的鸭肉糜来制备掺假羊肉,采用近红外漫反射方式在全波段范围(100004000 cm-1)内采集羊肉、掺假羊肉和鸭肉的近红外光谱图,分别考察多元散射校正(Multiplicative scatter correction,MSC)、标准正态变量变换(Standard normal variate correction,SNV)、面积归一化(Area normalization)、标准化(Autoscale)、15点平滑处理(Smoothing)、一阶导数处理(1stderivative)的光谱预处理方法对所建支持向量机(nuSVM)判别模型的预测效果。结果显示,不同光谱预处理所建nu-SVM判别模型预测效果不同。其中,经标准化处理后所建的nu-SVM模型的预测能力最差为90.38%;15点平滑处理后所建nu-SVM模型的预测效果最好(96.15%),对建模集正确判别率为99.07%,对检验集正确判别率为96.15%;其余处理所建nu-SVM模型的判别能力介于二者之间。结果表明,采用近红外光谱技术结合15点平滑预处理后所建nu-SVM模型可以实现羊肉中的掺杂鸭肉的鉴别。 

     

    Abstract: In order to realize the nondestructive determination of the adulterated mutton,a quick method of Fourier transform near infrared( FT- NIR) spectroscopy was employed to analysis the adulteration of duck meat in minced mutton.In this work,the adulterated mutton were obtained by interfusing different percentage of duck meat,and the original spectra of mutton,adulterated mutton and duck meat in the wave- number range of 10000~4000 cm- 1were obtained by spread reflection NIR. Then the predicted effects of different support vector machine( SVM)discriminate models that established by multiplicative scatter correction( MSC),standard normal variate transformation( SNV),area normalization,autoscale,15 point smoothing and 1stderivative were evaluated. The results showed that predicted effect of different nu- SVM models built by different preprocessing methods was different.The model accuracy to discriminate mutton,duck meat adulterated into mutton and duck meat was the poorest( 90.38%) built by autoscale preprocessing method,and good prediction model was obtained using the 15 point smoothing preprocessing method and the accuracy of the training set and prediction set of the model were99.07% and 96.15%,respectively.In addition,the predicted effects of other models built by the four preprocessing methods were between the both. Therefore,the nu- SVM model built by NIR combined with 15 point smoothing preprocessing weve proved to be a feasible method to identify duck meat in minced mutton.

     

/

返回文章
返回