文献类型：期刊文献

中文题名：基于近红外光谱法的两种杂交杨木物理力学重要品质性状快速预测

英文题名：Rapid Prediction for Important Physical and Mechanical Quality Traits of Two Types of Hybrid Poplars Based on Near Infrared Spectroscopy

作者：陈丽颖[1,2] 贺嘉琪[1] 贾茹[1] 李卓林[1] 王玉荣[1]

第一作者：陈丽颖

机构：[1]中国林业科学研究院木材工业研究所,北京100091;[2]南京林业大学,江苏南京210037

年份：2025

卷号：45

期号：S1

起止页码：542-547

中文期刊名：光谱学与光谱分析

外文期刊名：Spectroscopy and Spectral Analysis

收录：;EI(收录号：20260720049714);北大核心:【北大核心2023】；

基金：国家自然科学基金项目(32471799)资助。

语种：中文

中文关键词：杂交杨;物理性质;力学性质;近红外光谱;光谱预处理

外文关键词：Hybrid poplar;Physical properties;Mechanical properties;Near-infrared spectroscopy;Spectral preprocessing

分类号：S781

摘要：杨木物理力学品质性状是杂交杨树用材林选育的重要指标。采用近红外光谱技术预测杨木品质性状,掌握不同种杂交杨木材性变化,对实现优良人工林杨树定向培育、杨树遗传及加工改良等方面有重要科学意义。以栽植于辽宁绥中的中辽1号和渤丰杨1号两种杂交杨木为实验材料,利用近红外光谱仪采集试件不同切面光谱数据,并根据国家标准获得相应的真值。对采集的光谱数据采用不同预处理方法完成光谱平滑及滤波后,用完全交互验证法建立实际值与光谱数据之间的偏最小二乘回归(PLS)校正模型和预测模型。研究结果表明,两种杂交杨木的近红外光谱峰型基本一致,但不同切面的光谱吸收强度差异较大,横切面吸光度最高。基于横切面光谱信息建立的物理力学品质性状近红外偏最小二乘模型效果较好,但针对不同的品质性状应采取不同的预处理方法。对气干密度和抗弯强度预测时,应选取一阶导数法与卷积平滑(S-G)法对光谱进行预处理,对抗弯弹性模量和顺纹抗压强度预测时,应选取多元散射校正法(MSC)与S-G法对光谱进行预处理。选取最佳采谱切面和预处理方法建立中辽1号、渤丰杨1号两个杨木气干密度、抗弯强度、抗弯弹性模量、顺纹抗压强度预测模型,其相关系数分别为0.59和0.74、0.59和0.71、0.61和0.68、0.70和0.76,均方根误差分别为0.028和0.020、5.198和4.730、0.517和0.405、1.340和2.105。本研究分析了不同品系、采谱切面及预处理方法对近红外预测模型准确性的影响,得到了杂交杨木物理力学重要品质性状的最佳建模方法,为杨木近红外光谱的无损、快速检测提供了一定的技术支撑。
The physical and mechanical quality traits of poplar wood are important in selection and breeding.Using near-infrared spectroscopy(NIRs)to predict poplar wood quality traits and master the changes of wood properties of different hybrid poplars imposes a scientific significance for realizing oriented cultivation of artificial forest of poplars,reasonable and value-added utilization of poplar timbers,genetic and processing improvement of poplars.This study selected two hybrid poplars(Populus euramericana CL.‘Bofeng 1’,Populus×canadensis cv.‘Zhongliao 1’)collected from a forest farm in Suizhong City,Liao'ning Province as experimental material.The NIRs was used to collect the spectral data of the different wood sections of test specimens,and the true values were obtained by the national standard.After the spectral data completed smoothing filtering by various pretreatment methods,the partial least squares(PLS)calibration and prediction models was established based on complete interactive verification method.The results showed that the near-infrared spectral peak shapes of the two hybrid poplars were basically consistent,but the spectral absorption intensity of different sections was quite different,where the absorbance of cross section was the highest.The near-infrared PLS models for physical and mechanical quality traits established based on cross-sectional spectral information exhibited better performance,however,different pretreatment methods should be adopted for different quality traits.First derivative and convolution smoothing were the most suitable pretreatment method for the air-dry density and modulus of rupture(MOR)prediction models.Multivariate scattering correction(MSC)and convolution smoothing were the most suitable pretreatment method for the modulus of elasticity(MOE)and compression strength parallel to grain(CSPG)prediction models.The correlation coefficient(R_(p))of the density,MOE,MOR and CSPG prediction models of‘Zhongliao 1’and‘Bofeng 1’established with best spectrum section and pretreatment methods are 0.59 and 0.74,0.59 and 0.71,0.61 and 0.68,0.70 and 0.76,while the root mean square errors(RMSE)of which are 0.028 and 0.020,5.198 and 4.730,0.517 and 0.405,1.340 and 2.105.In this study,the effects of different cultivars,spectral sections and pretreatment methods on the accuracy of NIRs prediction models were analyzed and the best modeling methods for important physical and mechanical quality traits of hybrid poplar were obtained,which provided some theoretical support for non-destructiveand rapid detection of poplar wood near-infrared spectroscopy.