基于近红外光谱与KPCA-SVM鉴别淡水鱼肉

Identification of freshwater fish species based on near infrared spectroscopy and KPCA-SVM method

为实现淡水鱼品种的快速鉴别,采用近红外光谱分析技术建立7种淡水鱼鲜肉的快速鉴别模型。试验采集了鲢、草鱼、乌鳢、鲫、鲤、青鱼、鳙7种淡水鱼共772个鲜鱼肉样品的近红外光谱数据,分别考察标准正态变换(standard normalized variate,SNV)、多元散射校正(multiplicative signal correction,MSC)的预处理方法及核主成分分析(kernel principal component analysis,KPCA)和主成分分析(principal component analysis,PCA)的特征提取方法对支持向量机(support vector machine,SVM)判别模型的影响。结果显示,经SNV预处理和KPCA提取特征变量后,对未知样品的整体正确判别率达到92.68%。因此,采用近红外光谱技术结合化学计量学方法所建SVM模型可以实现淡水鱼品种的快速鉴别。

To realize the rapid identification of freshwater fish species,near infrared reflectance spectroscopy was employed to establish the identification models of fish species.772 samples of 7 freshwater fish species (silver carp,grass carp,snakehead,crucian carp,common carp,black carp,bighead carp) were prepared to collect near infrared spectra data.The effects of preprocessing methods including standard normalized variate (SNV),multiple scattering correction (MSC) and the feature extraction methods including kernel principal component analysis (KPCA) and principal component analysis (PCA) on the discrimination models of support vector machine (SVM) were investigated,respectively.The results showed that the overall accuracy rate was 92.68% for the unknown sample after the SNV preprocessing and KPCA extraction of characteristic variables.Therefore,the SVM model constructed by near infrared spectroscopy combined with chemometric methods is feasible for rapid identification of freshwater fish species.