大豆叶绿素含量高光谱反演模型研究

Inverse model for estimating soybean chlorophyll concentration using in-situ collected canopy hyperspectral data

叶绿素是植物体进行光合作用、进行第一性生产的重要物质,能够间接反映植被的健康状况与光合能力,同时也能反映植被受环境胁迫后的生理状态。高光谱遥感为快速、大面积监测植被的叶绿素变化提供了可能。该研究实测了不同水肥耦合作用下,大豆冠层的高光谱反射率与叶绿素含量数据,对二者进行了相关分析;采用特定叶绿素敏感波段建立了植被指数叶绿素估算模型;最后采用相关系数较大的波段作为神经网络模型的输入变量进行了叶绿素含量的估算。经对比发现叶绿素A、B与光谱反射率在可见光与近红外波段的相关系数的变化趋势基本一致,在可见光谱波段呈负相关,近红外波段呈正相关,红边处相关系数由负变正。特定色素植被指数可以提高大豆叶绿素估算精度(R2>0.736),但是人工神经网络模型可以大大提高大豆叶绿素含量的估算水平,当隐藏层节点数为4时,R2大于0.94,随着隐藏层节点数的增加,R2可高达0.99,表明神经网络模型可以大大提升高光谱反演大豆叶绿素含量的能力。

Chlorophyll is substance in vegetation for photosynthesis, ultimately affecting the net primary production, which can also indicate the healthy condition of vegetation living in a stressed environment. Hyperspectral remote sensing can provide a possibility for quick and accurate estimation of vegetation chlorophyll concentration in large areas. Soybean canopy reflectance data collected with ASD spectroradiometers （350～1050 nm）, which were cultivated in water-fertilizer coupled control conditions, and chlorophyll content data were collected simultaneously. First, correlation between reflectance, derivative reflectance against chl-A and chl-B was conducted; second, RVI, RARSa and PSSRb regressed against chl-A and chl-B; and finally, ANN-BP was established for soybean chlorophyll concentration estimation, which had different nodes in hidden layers. It was found that soybean canopy reflectance shows a negative relationship with chl-A and chl-B, while it shows a positive relationship with chl-A and chl-B in near infrared region. Reflectance derivative has an intimate relationship with chl-A and chl-B in blue, green and red edge spectral region, with the maximum correlation coefficient in red edge region. Chlorophyll specified absorption vegetation index has intimate relationship with chl-A and chl-B, with regression determination coefficient R^2 greater than 0. 736. ANN-BP model can greatly improve soybean chlorophyll concentration estimation accuracy. Determination coefficient （R^2 = 0.94） obtained with four nodes in hidden layers, however, R^2 still can be improved with nodes in hidden layers increasing, and R^2 reached 0.98 with six nodes in hidden layers. By above analysis, it indicated that, ANN-BP model can be applied to in-situ collected hyperspectral data for vegetation chlorophyll content estimation with quite accurate prediction, and in the future, ANN-BP model still should be applied to hyperspectral data for other vegetation biophysical and biochemical parameters estimation.