基于高光谱技术的牛肉-猪肉掺假检测  

Quantitative Detection of Beef-pork Adulteration by Hyperspectral Imaging

作　　者：白亚斌[1] 刘友华[1] 丁崇毅 冯慧[2] 陈伟[1] 冯耀泽[1] 

机构地区：[1]华中农业大学工学院,湖北武汉430070 [2]华中农业大学信息学院

出　　处：《海南师范大学学报（自然科学版）》2015年第3期270-273,共4页Journal of Hainan Normal University（Natural Science）

基　　金：湖北省自然科学基金重点项目(ZRZ2015000186);中央高校基本科研业务费专项基金资助项目(2662015PY051);华中农业大学新进博士科研启动基金

摘　　要：利用高光谱技术建立一种牛肉-猪肉掺假的快速无损检测方法.通过对样本集进行划分、光谱预处理以及特征波长的选择,建立了全波长和选择波长的掺假定量检测模型.结果表明:基于基线校正光谱预处理所得的偏最小二乘回归(PLSR)预测模型最优,其交叉验证决定系数R2CV=0.97、预测决定系数R2P=0.98,均方根误差RMSECV=5.20%、RMSEP=4.23%.此外,用回归系数法优选5个波长点作为特征波长进行建模,所得模型与全波长模型相比,具有变量少、结构简单、性能更优等优点.The study aims to establish beef-pork adulteration detection model employing non-destructive testing technique of hyperspectral imaging. By data set partitioning, spectral preprocessing as well as wavelength selection, full wavelength and simplified adulteration detection models were established. The best full wavelength partial least-squares regression（PLSR） model was obtained when the spectra were preprocessed by baseline correction（ R2 CV =0.97,R2 P =0.98, RMSECV=5.20%, RMSEP=4.23%）; The model built on five chosen characteristic wavelengths using regression coefficient method was superior to previous full wavelength model with fewer variables, simpler structure and better performance.

关 键 词：牛肉 猪肉 高光谱 掺假 基线校正 偏最小二乘回归法 

分 类 号：O657.3[理学—分析化学]