近红外高光谱技术在亚美马褂木与其亲本木材鉴别中的应用  

Wood identification of Liriodendron sino-americanum and hybrid parents by near-infrared hyperspectral imaging spectroscopy

作　　者：陈哲楠 高捍东 薛晓明[2] 吴皓琪 苏子珊 CHEN Zhenan;GAO Handong;XUE Xiaoming;WU Haoqi;SU Zishan(College of Forestry,Nanjing Forestry University,Nanjing 210037,China;Department of Criminal Science and Technology,Nanjing Forest Police College,Nanjing 210023,China;College of Landscape Architecture,Nanjing Forestry University,Nanjing 210037,China)

机构地区：[1]南京林业大学林学院,南京210037 [2]南京森林警察学院刑事技术学院,210023 [3]南京林业大学风景园林学院,南京210037

出　　处：《林业工程学报》2023年第2期166-171,共6页Journal of Forestry Engineering

基　　金：江苏省自然科学基金面上项目(BK20181338);江苏省高校优秀科技创新团队资助(2019-29);江苏高校“青蓝工程”项目。

摘　　要：亚美马褂木(Liriodendron sion-americanum)是一种通过中国马褂木和北美鹅掌楸杂交培育获得的优良阔叶工业树种与园林绿化树种。相比于亲本,亚美马褂木具有生长更快、适应性更广、材性更优等特点。由于亚美马褂木和其亲本木材在外观上相似度较高,因此木材交易中常出现以次充好的问题,破坏了市场秩序。为解决亚美马褂木木材鉴别问题,降低木材鉴别误差,本试验基于近红外高光谱(NIR-HSI)技术,建立亚美马褂木和其亲本的木材快速无损鉴别方法。使用近红外高光谱仪采集了亚美马褂木、北美鹅掌楸和中国马褂木木材样品的光谱信息,建立偏最小二乘判别分析(PLS-DA)和支持向量机(SVM)2种判别模型,比较了连续投影算法(successive projections algorithm, SPA)、SG平滑(S-G smoothing)、多元散射校正(multiplicative scatter correction, MSC)、归一化数据(Normalize)和标准正态变量变换(standard normal variate, SNV)5种预处理方法对建模的影响。试验结果表明:亚美马褂木与其亲本木材研磨成1~3 mm粒径粉末后更适合进行基于NIR-HSI技术的分类鉴别,而粉末粒径≤1 mm时鉴别准确率会下降(无数据预处理时校正集准确率最高下降4.9%);采用SPA预处理建立的SVM模型效果较优,在提取少量特征波长减少建模计算成本的同时,对3种木材粗粉末样品鉴别准确率为100%(校正集)。基于近红外高光谱仪结合合适的数据处理算法可以实现对亚美马褂木和其亲本的准确鉴别分析,为高度相似木材或木材制品的快速鉴别提供新的途径。本次试验使用粒径1~3 mm的木材粉末样品达到了较高的鉴别精度,为进一步降低木材鉴别误差提供了新的思路。Hybrid tuliptree(Liriodendron sion-americanum) is a hybrid species obtained by crossbreeding Chinese tuliptree(L. chinese Sarg.) and tuliptree(L. tulipifera). Hybrid tuliptree has great wood attribute and been widely used in furniture and other fields. Compared with parental species, hybrid tuliptree displays faster growth rate, better wood features and wider adaptability. However, the morphological properties of the species of hybrid tuliptree, Chinese tuliptree, and tuliptree are highly similar, which are difficult to be distinguished by their structural characteristics. Since the hybrid tuliptree offers obvious advantages for wood features and growth speed, timber traders often use Chinese tuliptree as hybrid tuliptree for profiteering. In order to solve the problem of hybrid tuliptree wood identification and reduce the identification error, a rapid nondestructive wood identification method was developed for distinguishing the hybrid tulip tree and its parents using the near-infrared hyperspectral(NIR-HSI) technology. The spectral information of the hybrid tulip tree, tulip tree, and Chinese tulip tree were collected using NIR-HSI. The partial least squares discriminant analysis(PLS-DA) and support vector machine(SVM) models were developed. The successive projections algorithm(SPA), SG smoothing(S-G smoothing), multiplicative scattering correction(MSC), normalize data(Normalize), and standard normal variate(SNV) were compared in order to improve the performance of models. The results showed that the hybrid tulip tree and its parents were more suitable for identification using the NIR-HSI technique after grinding into powder with a diameter of 1-3 mm, while the identification accuracy would be reduced when powder diameter≤1 mm(up to 4.9% drop in calibration set accuracy with no data pre-processing). After the SPA treatment, the established SVM model can achieve 100% accuracy(calibration set) by rough grinding powder. This method has high accuracy while reducing the computational cost of modeling by extracting

关 键 词：亚美马褂木 亲本树种 木材鉴别 近红外高光谱 

分 类 号：S794[农业科学—林木遗传育种] Q917[农业科学—林学]