木材红外光谱的树种识别研究  

Research on Infrared Spectrum for Timber Species Identification

作　　者：王学顺[1] 孙一丹[1] 黄安民[2] 

机构地区：[1]北京林业大学理学院,北京100083 [2]中国林业科学研究院木材工业研究所,北京100091

出　　处：《森林工程》2015年第6期65-70,共6页Forest Engineering

基　　金：国家自然科学基金(31270591)

摘　　要：以10种珍贵木材30个样本的红外光谱（1800~800 cm-1）为研究对象,分别建立了基于红外光谱的木材树种定性与定量识别模型。在定性分析中,选取5种木材的15个样本的红外光谱,通过主成分分析进行光谱数据降维,并利用主成分投影判别法分别得到其二维和三维主成分得分图,对样本进行直观分类;在定量分析中,分别建立了木材红外光谱的聚类分析模型、贝叶斯判别模型以及支持向量机模型,聚类分析与贝叶斯判别模型木材判别准确率分别为83.33%和86.67%。在支持向量机模型中,分别采用网格搜索法与遗传算法对支持向量机模型进行参数寻优,木材判别准确率分别为86.67%和85%。结果表明,利用木材红外光谱可以对木材树种进行有效识别,本研究为红外光谱技术在森林工程领域的应用提供一定的科学依据与参考价值。A total of 30 infrared spectra of samples of 10 kinds of precious wood were selected as the research objects,and the model of timber species identification was established around the qualitative analysis and quantitative analysis of infrared spectrum. In the qualitative analysis,15 infrared spectra of samples of 5 kinds of wood were selected for data dimension reduction by principal component analysis,and two dimensional principal component score plot as well as three dimensional principal component score plot was drawn respectively for timber intuitive classification. In the quantitative analysis,the models of clustering analysis,Bayes discriminant,and support vector machine were established,respectively. The discriminant accuracy rate of clustering analysis and Bayes discriminant model was 83. 33% and 86. 67%,respectively. In support vector machine（ SVM） model,the discriminant accuracy of parameter optimization based on grid search method and genetic algorithm reached 86. 67% and 86. 67%,respectively. The results indicated that infrared spectra was effective for timber identification,which provided a certain reference value in the application of forest engineering.

关 键 词：木材识别 红外光谱 主成分分析 支持向量机 

分 类 号：S781.82[农业科学—木材科学与技术] TS612.7[农业科学—林学]