基于变量优选和近红外光谱技术的红富士苹果产地溯源

Tracing the origin of Red Fuji apple based on variable optimization and near-infrared spectroscopy

为实现对红富士苹果的产地溯源,采集阿克苏、静宁、灵宝、烟台的红富士苹果近红外光谱数据,分别采用归一化、中心化、一阶导数、二阶导数、标准正态变换、多元散射校正(multivariate scattering correction,MSC)、小波变换、SG平滑变换、傅里叶变换等9种方法对原始光谱进行预处理,建立概率神经网络(probabilistic neural network,PNN)模型对苹果的产地进行识别。结果表明,MSC预处理之后的模型总准确率最高,为97.5%,阿克苏、静宁、灵宝、烟台4个产地的准确率分别为100%、100%、90%、100%。为简化模型,对MSC预处理之后的光谱数据分别采用主成分法、连续投影算法(successive projection algorithm,SPA)、竞争性自适应重加权算法(competitive adaptive reweighted sampling,CARS)、随机蛙跳算法(random frog,RF)、CARS-SPA、RF-SPA选取特征变量建模。综合考虑正确率和模型的复杂性,最优模型MSC-CARS-SPA-PNN的测试集的总准确率为98.75%,4个产地的红富士苹果准确率分别达到了100%、100%、95%、100%。该研究可为红富士苹果的产地溯源提供理论参考。

Near-infrared spectrum data of Red Fuji apples from Aksu,Jingning,Lingbao,and Yantai were collected to trace the origin of Red Fuji apples.Nine methods including normalization(NOR),centralization(CEN),first derivative(1-DER),second derivative(2-DER),standard normal transform(SNV),multivariate scattering correction(MSC),wavelet transform(WT),SG smoothing transform(SG),and Fourier transform(FT)were used to preprocess the original spectrum.Results showed that the model after multivariate scattering correction pretreatment had the highest recognition rate of 97.5%,and the recognition rates of Aksu,Jingning,Lingbao,and Yantai were 100%,100%,90%,and 100%,respectively.To simplify the model,principal component analysis(PCA),successive projection algorithm(SPA),competitive adaptive reweighted sampling(CARS),random frog(RF),and their combination algorithms were used to select characteristic variables.Results showed that the total recognition rate of MSC-CARS-SPA-PNN was 98.75%,and the recognition rates of Red Fuji apples from four producing areas were 100%,100%,95%,and 100%,respectively,which could provide theoretical reference for the origin discrimination of Red Fuji apples.