机构地区: 华南农业大学工程学院
出 处: 《湖南科技学院学报》 2011年第8期36-39,共4页
摘 要: 水分含量是衡量砂糖橘营养品质的重要指标之一,其快速无损检测显得越来越重要。本文基于小波变换的方法,对砂糖橘的500-2500nm区间的漫反射光谱,利用正交小波函数DBn(n=2,3,…10)分别进行2-6五个水平分解和消噪,并比较了不同小波函数和不同分解水平的消噪效果。结果表明,小波消噪有利于消除导数光谱中的噪声,提高建模精度,基于小波函数DB3(分解尺度为3)消噪后的导数光谱建立的PLS模型的预测相关系数为0.8725,预测均方根误差为0.6767。 water content is one of the most important nutrition material in shatangju oranges.The fast and nondestructive examination of water content in sugar oranges is more and more important.Firstly,the spectra of Shatangju samples within 500-2500nm were decomposed in level 2-6 using the orthogonal wavelet functions DBn(n=2,3,…10).And then the de-noise results of different wavelet functions and different decomposing levels were compared.Wavelet de-noise(WD) was examined to be the optimal spectrum preprocessing method.The PLS model based on derivative spectra with DB3 wavelet de-noise(in decomposition level 3) produced best result(RMSEP=0.6767,Rp =0.8725).