近红外光谱结合支持向量机快速识别树种

利用阿达玛变换近红外光谱结合支持向量机,对制浆造纸常用木材树种的快速识别进行研究。将各树种近红外光谱先进行多点平滑和标准正态变换预处理以消除噪音干扰和光散射导致的测量偏差,然后基于不同建模策略建立一对多和一对一两种支持向量机模型,考察这两种模型对多树种属间分类和种间分类的预测能力,并与传统的偏最小二乘判别分析分类法进行对比。结果表明,支持向量机预测模型对桉木、相思木、杨木、水杉等树种的属间分类正确率达到98%以上,种间分类正确率均达到95%以上,在处理复杂分类问题时模型稳健性明显优于传统分类方法,从方法上证明了近红外技术工业化应用的可能性,为进一步建立近红外在线检测木片材性分析系统奠定了基础。

Fast Identification of Wood Species by Near Infrared Spectroscopy Coupling with Support Vector Machine

Fast identification of different wood materials for papermaking by portable hadamard transform near infrared spectroscopy (HT-NIR) in combination with support vector machines (SVM) was investigated in present study. Savitzky-Golay smoothing method and standard normal variation were used to pretreat the spectral for eliminating noise and measurement deviation caused by light scattering. The one-against-all model and one-against-one model were constructed based on different SVM modeling strategies. The prediction performance for genera classification and species classification of two SVM models was evaluated with partial least squares discriminant analysis (PLS-DA). In this study,SVM was applied to identify different wood species, such as eucalyptus, acacia, populus and metasequoia. The genera correct classification rates and species correct classification rates achieved above 98% and 95%, respectively. The SVM method demonstrated its integrated merits in solving complex classification compared with the traditional linear machine learning methods. The study results showed the feasibility of industrial application of NIR technology and laid the foundation for building the on-line NIR analysis system for wood chips.