关键词：
同步荧光法;最小二乘支持向量回归;金霉素;小波;鸭肉
摘 要：
为实现鸭肉中金霉素残留含量的快速测定,提出了应用同步荧光法结合最小二乘支持向量回归(LSSVR)来建立鸭肉中金霉素残留含量的定量分析模型。首先应用三维同步荧光法对波长250～450 nm的同步荧光光谱进行分析,确定检测鸭肉中的金霉素含量的最佳波长差Δλ为70 nm;然后使用sym8小波的2层分解对原始同步荧光光谱进行光谱预处理,对预处理后的光谱采用遗传算法(GA)结合交互验证均方根误差(RMSECV)方法选择了16个波长作为检测模型的输入特征向量;最后对LSSVR、BP和SVM_SteveGunn 3种模型进行性能比较,以LSSVR模型的预测效果最好,其预测集的决定系数R2和预测均方根误差分别为0.9491和2.5660。试验结果表明,同步荧光法结合LSSVR模型检测鸭肉中的金霉素残留含量是可行的。
译 名：
Determination of Aureomycin Content in Duck Meat Using LSSVR and Synchronous Fluorescence Spectroscopy
作 者：
Zhao Jinhui Yuan Haichao Liu Muhua Xiao Haibin Xu Jiang(College of Engineering,Jiangxi Agricultural University,Nanchang 330045,China)
关键词：
Synchronous fluorescence spectroscopy,LSSVR,Aureomycin,Wavelet,Duck meat
摘 要：
In order to achieve the rapid prediction of aureomycin contents in duck meat,quantitative analysis models of aureomycin contents in duck meat were established by using synchronous fluorescence spectroscopy and LSSVR method.Firstly,synchronous fluorescence spectroscopies between 250 nm and 450 nm were analyzed by using 3-D synchronous fluorescence spectroscopy,and 70 nm was selected as the optimum wavelength difference.Secondly,the sym8 wavelet with two levels decomposition was used to complete the pretreatment of the original synchronous fluorescence spectroscopy,and 16 characteristic wavelengths were selected as the inputs of detection model combining with genetic algorithm and the root mean square error of cross-validation.Lastly,the performances of three models of LSSVR,BP and SVM_SteveGunn were compared.LSSVR model was optimal among them,and the determination coefficient R2 and the root mean squared error of prediction samples were 0.949 1 and 2.566 0,respectively.The experimental results showed that it was feasible to predict the aureomycin contents in duck meat combining with synchronous fluorescence spectroscopy and LSSVR.
