基于近红外光谱法的木材原料的Kohonen网络分类法

Kohonen Network Discrimination of Wood Materials Based on NIR Spectroscopy

使用近红外光谱法结合Kohonen网络对不同的造纸木材原料进行了快速分类与评价。先对原料的近红外光谱数据进行小波变换预处理,再利用主成分分析法对小波变换预处理后的光谱数据进行信息的压缩和提取,选择了第一、二主成分可表达原有光谱数据信息,再把这2个主成分输入Kohonen网络进行分类处理,从处理结果来看可较好地区分马尾松、杨木和桉木等3种不同制浆性能的木材原料,且能较好表达木材原料类别之间的性质关系。

The classification of different wood materials for papermaking based on near-infrared spectroscopy and Kohonen network was carried out. The spectroscopic data pretreated by wavelet transformation was extracted into two principal components with principal components, analysis. The two principal components, data represented the spectroscopic information of the materials were input into a Kohonen network to classify the wood materials for pa- permaking. The results showed that the network could not only exactly discriminate the species of Masson＇s pine, poplar and eucalyptus, but also elucidate their qualities＇ connections of these classes.