近红外结合极限学习机快速识别牛肉中掺假猪肉  

Rapid Identification of Beef Adulteration with Pork by Near-infrared Combined Limit Learning Machine

作　　者：韩方凯[1] 刘璨 黄煜 马梅 冯凡 段腾飞[1] 张东京[1] HAN Fang-kai;LIU Can;HUANG Yu(Suzhou University,Suzhou,Anhui 234000)

机构地区：[1]宿州学院,安徽宿州234000 [2]江苏大学,江苏镇江212013

出　　处：《安徽农业科学》2019年第13期185-187,共3页Journal of Anhui Agricultural Sciences

基　　金：安徽省自然科学基金项目(1908085QC146);安徽高校自然科学研究重点项目(KJ2018A0447);宿州学院教授(博士)科研启动基金项目(2016jb03)

摘　　要：[目的]建立牛肉中掺假猪肉的快速鉴定方法。[方法]采用傅立叶变换近红外结合极限学习机(extreme learning machine,ELM)构建纯牛肉、牛肉中掺假猪肉、纯猪肉的快速识别模型,考察标准正态变换、多元散射校正(multiplicative scatter correction,MSC)、一阶导数及二阶导数结合核主成分分析(kernel principal componentan alysis,KPCA)等光谱预处理方法对ELM模型预测性能的影响。[结果]MSC+KPCA预处理下,ELM模型的预测效果最优,训练集及测试集的正确识别率分别为86.67%和83.33%。[结论]近红外光谱技术结合ELM在牛肉中掺假猪肉的快速鉴定方面具有较大的潜力。[Objective]The research aimed to develop a rapid technique for identification of beef adulteration with pork.[Method]The Fourier transform near-infrared combined with extreme learning machine (ELM) was used to build prediction models for identification of pure beef, beef adulteration with pork, and pure pork. The influence of different spectral pretreatment methods on the performance of ELM models were studied, such as standard normal variate transformation, multiplicative scatter correction (MSC), first derivative and second derivative combined separately with kernel principal component analysis (KPCA).[Result]The best ELM model was obtained under MSC+KPCA with the correct recognition rate in train set and prediction set was 86.67% and 83.33% respectively.[Conclusion]The Fourier transform near-infrared in coupled with ELM has a great potential in rapid identification of beef adulteration with pork.

关 键 词：牛肉掺假 近红外 极限学习机 核主成分分析 

分 类 号：TS207.3[轻工技术与工程—食品科学]