基于NIRS与支持向量机的落叶松木材密度预测

Modeling Wood Density with NIRS and Support Vector Machine for Dahurian Larch

在支持向量机的理论基础上,以117个落叶松样本作为实验材料,用常规方法测定样本的密度实值,用美国ASD公司生产的波长为350-2500 nm的Lab Spec近红外光谱仪对样本进行相应的光谱采集,对光谱数据进行预处理,以文本格式导出。用LibSVM在matlab环境下建立落叶松密度预测模型。经分析,该模型对训练集的回归拟合,R^2达到了85.04%,均方差为6.46×10^-4;对测试集的回归拟合,R^2为85.20%,均方差为4.45×10^-4,拟合效果较好。结果表明,该方法可以用于落叶松木材密度预测。

In this study,based on support vector machine（ SVM） theory,the sample data was based on 117 samples of dahurian larch and the wood density was determined by conventional methods of real value,NIR were collected from dahurian larch with LabSpec near infrared spectrometer which wavelength is 350 - 2500 nm of ASD,and built of dahurian larch density of NIR prediction models in matlab environment after preprocessing the spectral data and exporting in text format. Through the analysis,the model of training set regression fitting,R^2 reached 85. 04% and mean square error of 6. 46 × 10^- 4; Regression fitting of test set,R^2was85. 20% and the mean square error of 4. 45 × 10^- 4,better fitting effect,Research shows that this method can be used for larch wood density prediction.