摘要
基于小波变换的原理,分别利用阈值滤波、小波包、小波收缩3种常用的去噪方法,对砂糖橘样品的可见/近红外光谱信号进行去噪处理,探讨每种去噪方法的最优参数组合(小波函数、分解尺度、阈值)的同时选择最适去噪方法,并通过偏最小二乘法(PLS)对去噪后的重构光谱和砂糖橘果形指数建模。结果表明,小波包去噪有利于消除导数光谱中的噪声,提高建模精度,其最优参数组合为默认阈值条件下,小波函数Bior1.3、2尺度分解,去噪后的砂糖橘果形指数光谱建立的PLS模型的预测集R为0.9632,RMSEP为0.0779。
The visible / near infrared spectroscopy signals in citrus shatangju samples were de-noised by threshold filtering,wavelet packet,wavelet shrinkage,respectively,based on the principle of wavelet transform. The fruit shape index determination model was established by partial least squares（ PLS） method. Results showed that wavelet packet can remove the noise effectively from the derivative spectrum,and improve the accuracy of the prediction model. The optimal parameter combination was the Bior 1. 3 wavelet function,two levels of decomposition,with default threshold selection rule. The prediction correlation coefficient（ R） of PLS model of the titled index was 0. 9632,and the standard error of prediction（ RMSEP） was 0. 0779.
出处
《山西农业大学学报（自然科学版）》
CAS
2014年第6期558-563,共6页
Journal of Shanxi Agricultural University(Natural Science Edition)
基金
国家十二五科技支撑计划项目(2012BAD38B07)
关键词
砂糖橘
可见/近红外光谱
去噪
果形指数
Citrus Shatangju
Visible / near infrared spectroscopy
De-noising
Fruit shape index
