玉米叶片叶绿素含量高光谱反演模型探究.docVIP

玉米叶片叶绿素含量高光谱反演模型探究 　　摘要：叶绿素含量是绿色植物生长状态的一个重要指标。首先在实验室采集玉米叶片高光谱数据和测定叶绿素含量，并对光谱数据进行对数一阶微分变换，对比选取建模反演因子。根据选定的反演因子采用线性回归、模糊识别和BP神经网络方法建立了玉米叶片叶绿素含量高光谱反演模型，并计算出模型的精度。结果表明，有较好非线性映射能力的BP神经网络反演模型能够高精度地反演出玉米叶片中的叶绿素含量。BP神经网络模型叶绿素含量预测和实测叶绿素含量的平均绝对误差（e）为1.126，决定系数（R2）为0.902，均方根误差（RMSE）为1.375。玉米叶片叶绿素含量与高光谱数据并非线性关系，BP神经网络反演模型能够较好地运用到叶片叶绿素含量反演中。 　　关键词：玉米叶片；叶绿素含量；高光谱遥感；反演模型 　　中图分类号：P237；S513 文献标识码：A 文章编号：0439-8114（2015）11-2744-05 　　DOI：10.14088/j.cnki.issn0439-8114.2015.11.049 　　Researches on Hyperspectral Inversion Model of Corn Leaf Chlorophyll Content 　　YANG Ke-ming， SUN Yang-yang，WANG Lin-wei， SHI Gang-qiang， WEI Hua-feng， LIU Fei 　　（College of Geosciences and Survey Engineering， China University of Mining Technology （Beijing）， Beijing 100083， China） 　　Abstract：Chlorophyll content is an important indicator of green plant growth status. Firstly the hyperspectral data and chlorophyll content of corn leaf were measured in the laboratory， and the inversion factors were selected by comparing the data which was transferred by the hyperspectral data of corn leaf through the first order differential of logarithmic. Then hyperspectral inversion models of corn leaf chlorophyll content were established by method of liner regression， BP neural network and fuzzy recognition based on the selected inversion factors and the accuracy of each model was calculated. Results showed that the BP neural network inversion model with good nonlinear mapping capability could accurately predict the chlorophyll content of corn leaf high. The mean absolute error was 1.126，R2 was 0.902 and RMSE was 1.375 in BP neural network inversion model. The chlorophyll content and hyperspectral data of corn leaf was not a linear relationship and the BP neural network inversion model could be applied in leaf chlorophyll inversion. 　　Key words：corn leaf； chlorophyll content； hyperspectral remote sensing； inversion model 　　高光谱遥感技术是21世纪遥感技术重大、影响深远的突破，在国民经济发展中发挥了重要作用[1]。高光谱波谱分辨率高，光谱连续性强，蕴含着丰富的地物信息，为遥感从对地定性观测到定量观测提供了