结合WT预处理的近红外光谱PLS算法预测鲜枣糖度

[目的]研究结合WT预处理的近红外光谱PLS算法模型预测鲜枣糖度的方法。[方法]用S-G、MSC、FD、SD、WT和WT＋MSC 6种预处理法,SMLR、PCR和PLS 3种算法模型,对60个鲜枣样品的近红外光谱数据进行预处理、糖度预测和建模精度分析,建立最佳算法的数学模型。[结果]在鲜枣糖度近红外光谱预处理阶段引进小波变换方法去除导数光谱噪声,得到了很好的去噪效果。不同的小波函数、分解尺度使消噪的结果有所不同。与常见的光谱预处理法相比,在选用db4-3小波函数、默认阈值情况下,采用WT＋MSC预处理及建模算法为PLS时所建立的模型最好,其相关系数R为0.919 02,校正集标准差RMSEC为0.863,预测集标准差RMSEP为1.71。[结论]结合小波变换预处理的PLS算法模型可有效预测鲜枣糖度,改善模型的预测精度。

Prediction on Sugar in Fresh Jujube Based on Near Infrared Spectroscopy PLS Algorithm Combining with WT Pretreatment

[Objective]The aim was to study the method of predicting the sugar in the fresh jujube based on near infrared spectroscopy PLS algorithm models combining with the wavelet transform(WT) pretreatment.[Method]The near infrared spectroscopy data of 60 fresh jujube samples were made for the pretreatment,sugar prediction and modeling accuracy analysis by using 6 pretreatments including S-G,MSC,FD,SD,WT and WT+MSC and 3 algorithm models including SMLR,PCR and PLS and the optimum algorithm models was established.[R...