基于高光谱分析技术的桉树叶片黄化识别  

Yellowing Identification in Eucalypt Leaves Based on Hyperspectral Analysis Technology

作　　者：赵隽宇 石媛媛[1] 杨瑞青 邓昀[3] 程小辉[3] 陈守学[3] 曹继钊[1] 唐健[1] Zhao Junyu;Shi Yuanyuan;Yang Ruiqing;Deng Yun;Cheng Xiaohui;Chen Shouxue;Cao Jizhao;Tang Jian(Guangxi Forestry Research Institute,Guangxi Research and Development Center for New Forestry Fertilizer,Nanning,Guangxi 530002,China;Guangxi Huawote Group Co.,Ltd.,Nanning,Guangxi 530025,China;Guilin University of Technology,Guilin,Guangxi 541004,China)

机构地区：[1]广西壮族自治区林业科学研究院、广西林用新型肥料研发中心,广西南宁530002 [2]广西华沃特集团股份有限公司,广西南宁530025 [3]桂林理工大学,广西桂林541004

出　　处：《广西林业科学》2024年第1期94-99,共6页Guangxi Forestry Science

基　　金：中央财政林业科技推广示范项目([2021]TG18号);中央引导地方科技发展资金项目(桂科ZY21195026);广西林业科技推广示范项目(桂林科研[2021]24号)。

摘　　要：桉树黄化病是一种较特殊的生理性病害,存在一定的突发性和随机性,及时发现并补充养分可大幅减少病害带来的损失。采集桉树黄化叶片、未发病叶片和正常叶片,采用高光谱仪测定不同叶片的光谱特征,基于偏最小二乘法判别分析(Partial Least Squares Discriminant Analysis,PLS-DA)和正交偏最小二乘法判别分析(Orthogonal Partial Least Squares Dis-criminant Analysis,OPLS-DA)方法,分别建立判别分析模型,对比模型判别效果。结果表明,不同叶片光谱反射曲线呈相同趋势,反射率差异明显,差异较大的波段主要为近红外波段800~1260、1400~1720和2000~2400 nm,受病害影响叶片的原始光谱反射率明显高于正常叶片;对数变换可在一定程度上减少光谱数据冗余量,突出差异;两种线性判别分析方法均能识别潜在黄化叶片,Log-OPLS-DA的判别效果更好,模型R2为0.91,RMSE为0.203。高光谱分析技术结合OPLS-DA对桉树黄化叶片具有一定的预测和识别潜力。Eucalypt yellowing disease is a relatively specific physiological disease,which has certain degree of suddenness and randomness.Timely detection and supplementation of nutrients could significantly reduce losses caused by the disease.Yellowing leaves(Chlorosis),non-diseased leaves(Chlorosis-Normal)and healthy leaves(Normal)of eucalypt were collected,and spectral characteristics of different leaves were determined by hyperspectral instrumentation.Discriminant analysis models were established based on PLS-DA and OPLSDA methods,and discriminant effects were compared.Results showed that spectral reflection curves of different leaves showed the same trend,but reflection rates had obvious differences.Bands with significant differences were mainly reflected in near-infrared bands of 800-1260,1400-1720 and 2000-2400 nm.Original spectral reflection rates of leaves affected by disease were significantly higher than those of normal leaves.Logarithmic transformation could reduce redundancy of spectral data to a certain extent and highlight differences.Both linear discriminant analysis methods could identify potential yellowing leaves.Discriminant effect of Log-OPLS-DA was better with model R2 of 0.91 and RMSE of 0.203.Combination of hyperspectral analysis technology and OPLS-DA had certain potentials of prediction and determination for eucalypt yellowing leaves.

关 键 词：高光谱 黄化叶片 无损检测 病虫害防治 桉树人工林 

分 类 号：S792.39[农业科学—林木遗传育种]