基于可见/近红外光谱识别东北地区常见木材

探索可见/近红外光谱识别生长锥取得的木材样品的可行性，为野外木材检测提供方法。木材品种涉及东北地区常见的14个树种，木材样品由生长锥南向北方向钻取树木1.3 m胸高处得到，观测木样的可见/近红外光谱并采用导数、对数与平滑处理，运用距离法建立识别模型，测试并分析不同光谱预处理方法对可见/近红外光谱木材识别准确率的影响。结果显示:在不使用平滑处理时，使用一阶导数处理之后的光谱木材树种识别的准确率（96.79%）明显高于二阶导数（78.57%）和三阶导数（75.00%）。在使用导数和平滑处理时，使用二阶导数（98.21%）或三阶导数（98.21%）处理之后的光谱用于识别的准确率略高于一阶导数（97.50%）。单独使用平滑处理不能提高模型预测准确率，单独使用导数处理可以提高模型预测准确率。在最优的参数设置下使用导数和平滑处理时，使用S-G导数平滑（98.42%）和Norris导数滤波（98.57%）的效果无明显差异。

Identification of common wood species in northeast China using Vis/NIR spectroscopy

To determine the feasibility of using visible/near infrared spectroscopy to identify wood samples drilled using an increment borer, and to provide a method of wood identification in the wild, 14 wood species typically found in Northeast China were sampled and drilled with an increment borer through 1.3 m breast-height of the trees from south to north, the samples are about 300 mm in length and 5 mm in diameter. Analysis included use of derivative, logarithmic, and smooth processing to process the wood spectrum and used a distance method to build the identification model. Results showed that without smoothing, the accuracy rate of the first derivative (96.79%) was much better than the second derivative (78.57%) and the third derivative (75.00%). With derivative and smoothing, the accuracy rate of the second derivative + smoothing (98.21%) and the third derivative + smoothing (98.21%) were better than the first derivative + smoothing (97.50%). Accuracy rate of the prediction model was not improved with just smoothing, and the derivative pretreatment improved accuracy. Also, with optimum parameters, there were no major differences between S-G derivative smoothing (98.42%) and the Norris derivative filter (98.57%). Thus, this study provided a new method and idea for rapid identification of wood species.