基于高光谱的苹果叶片叶绿素含量估算

Estimation of apple leaf chlorophyll content based on hyperspectral data

以2012、2013年山东省肥城市潮泉镇下寨村的苹果叶片为研究对象,分析叶片叶绿素含量与原始光谱反射率、连续统去除光谱之间的相关性,探索苹果叶片叶绿素含量的估算模型。结果显示:苹果叶片叶绿素含量与原始光谱相关性最好的波段在553、711和1 301 nm处,其中,以711 nm处的光谱所建立的模型最佳(R2=0.88);与连续统去除光谱相关性最好的波段在553、738和801 nm处,其中,以738 nm处的光谱所建立的模型最佳(R2=0.94)。根据相关性所选的敏感波段,利用随机森林(random forest,RF)建立基于以上6个波段的叶绿素含量预测模型(R2=0.94)。对所建立的711 nm、738 nm、RF算法估算模型进行检验,结果表明,利用RF建立的苹果叶片叶绿素含量模型最佳(R2=0.54)。

In the present study,the apple leaf spectral and chlorophyll content of apple leaf were acquired from 2012 to 2013 in Feicheng City,Shandong Province. The correlation between chlorophyll content and the original spectral reflectance or continuum removal spectrum was analyzed to explore the estimation model of apple leaf chlorophyll content. It was shown that the optimal band correlation between apple leaf chlorophyll content and the original spectral reflectance was 553,711 and 1 301 nm,and the optimal model for apple leaf chlorophyll content was obtained based on 711 nm spectrum,of which the determination coefficient was 0. 88. The optimal band correlation between apple leaf chlorophyll content and continuum removal spectrum was 553,738 and 801 nm,and the optimal model of apple leaf chlorophyll content was obtained based on 738 nm spectrum,of which the determination coefficient was 0. 94.According to the sensitive wavelength based on correlation,the apple leaf chlorophyll content prediction model was established based on random forest（ RF）,of which the determination coefficient was 0. 94. The established estimation models for apple leaves chlorophyll content based on random forest,spectral parameters 711 and 738 nm were compared,and it was shown that the optimum model was established by random forest,of which the determination coefficient was 0. 54.