联合时序遥感影像和地基激光雷达的玉米生长过程监测方法研究

Corn growth process monitoring is very important to monitor growth state and its change, potential stress and yield estimation. There are several limitations in the field of corn growth process monitoring currently, such as a lack of considering the leaves within the canopy, lower spatial resolution and bad temporal/spatial continuity of monitored results. Aiming at resolve these problems, we do the growth process monitoring of maize by combining time series spectral remote sensing images and terrestrial laser scanning data. We will develop multi-layer PROSAIL radiative transfer model to consider the leaves within the canopy. We optimize the probability density distribution function of leaf inclination angle of multi-layer PROSAIL model by calculating the leaf inclination angle of each horizontal layer. We retrieve continuous corn growth state parameters spatially and temporally by blending MODIS and Landsat 8 OLI remote sensing images based on Kalman Filter-based method, including Leaf Area Index, leaf inclination angle, leaf chlorophyll content, with considering the vertical difference within corn canopy. At last, we develop corn growth state monitoring model and do the potential stress diagnosis analysis of corn growth. This project aims at improve the accuracy and precision of corn growth process monitoring by combining time series spectral remote sensing images and terrestrial laser scanning data, considering the leaves within the canopy using multi-layer PROSAIL model.

玉米生长过程监测对于了解其生长状况、潜在胁迫、估产具有重要意义。针对目前的监测方法对冠层中、下层叶片信息考虑有限、监测结果的分辨率和时空连续性有待加强的问题，本项目联合使用地基激光雷达数据和时序遥感影像，挖掘中、下层叶片信息，构建垂直分层的PROSAIL辐射传输模型；利用卡尔曼滤波递推方法对MODIS、Landsat 8影像进行时空尺度转换，从而提高影像的空间分辨率并保证其时空连续性；基于地基激光雷达点云数据提取玉米冠层各水平层的叶倾角，优化垂直分层PROSAIL模型的叶倾角概率密度分布函数，并采用高斯处理器法反演得到整个生长季内时空连续的、垂直分层的玉米冠层叶面积指数、叶倾角、叶绿素含量；构建玉米生长过程监测模型，进行潜在胁迫诊断分析。本项目的特色和创新是联合激光雷达和时序遥感影像，在构建垂直分层PROSAIL模型、考虑中下层叶片信息的基础上，提高玉米生长过程监测精度和精细水平。

在玉米生长过程监测过程中，存在冠层参数反演时对冠层中下层叶片信息考虑有限、高温多雨导致在玉米生长内可用的光学影像不足等问题。本项目以河北省保定市的玉米种植区为研究区，（1）联合使用激光雷达数据和MODIS、Sentinel-2光学遥感影像，构建垂直分层的PROSAIL辐射传输模型，挖掘玉米冠层中下层叶片信息。研究结果表明：玉米叶倾角在垂直方向上具有明显的分布规律，上层叶片较为直立，中、下层叶片较平展，所构建的垂直分层PROSAIL模型考虑了叶倾角的垂直特征及生长变化。（2）研发卡尔曼滤波方法，挖掘时间序列MODIS影像的时间轨迹和Sentinel-2影像的地物空间细节，构建高时空分辨率遥感影像。研究结果表明：卡尔曼滤波算法能够稳定、准确的融合生产长时间序列的反射率影像，且影像兼具MODIS影像的高时间分辨和Sentinel-2影像的高空间分辨率，能够弥补玉米生长季内因云、雨影响导致光学遥感影像缺失的问题，是全生育玉米冠层参数反演的理想数据源。（3）基于地基激光雷达点云数据提取玉米冠层各水平层的叶倾角，优化垂直分层PROSAIL模型的叶倾角概率密度分布函数，并采用高斯处理器法反演得到整个生长季内时空连续的、垂直分层的玉米冠层叶面积指数、叶倾角、叶绿素含量。研究结果表明：利用改进的玉米叶倾角概率密度分布函数，将玉米冠层LAI反演精度从R2为0.76（p<0.001）提高到R2为0.82（p<0.001），同时将RMSE从0.56 m2/m2降低到0.31 m2/m2。