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800～1 260、1 400～1 720和2 000～2 400 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.
Yellowing Identification in Eucalypt Leaves Based on Hyperspectral Analysis Technology
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 loss-es 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 OPLS-DA 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 800-1 260,1 400-1 720 and 2 000-2 400 nm.Original spec-tral reflection rates of leaves affected by disease were significantly higher than those of normal leaves.Logarith-mic 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 technol-ogy and OPLS-DA had certain potentials of prediction and determination for eucalypt yellowing leaves.
国家科技期刊平台
NSTL
万方数据
