When the remote sensing pixel is composed of multiple components and a non-isothermal surface, its directional signature of thermal-infrared radiation is mainly determined by the 3D structure of the pixel. In this pap...When the remote sensing pixel is composed of multiple components and a non-isothermal surface, its directional signature of thermal-infrared radiation is mainly determined by the 3D structure of the pixel. In this paper, we present our simple directional thermal radiation model to describe the relation between the pixel thermal emission and the pixel’s component parameters, and invert the model to get the component temperatures. For the inversion algorithm, we focus on how to use the information of given observations in a more effective way. The information content in data space and parameter space is defined, and the transferring of information content in inversion procedure is studied. Our forward model and inversion method are validated using indoor directional measurement data.展开更多
通过无人机多光谱影像反演农作物理化参数、动态监测作物长势是精准农业发展的重要方向。然而,由于无人机影像多具有较高的空间分辨率,地面采样点与影像上对应像素的空间范围往往不匹配,导致所构建的反演模型精度降低。为确定利用无人...通过无人机多光谱影像反演农作物理化参数、动态监测作物长势是精准农业发展的重要方向。然而,由于无人机影像多具有较高的空间分辨率,地面采样点与影像上对应像素的空间范围往往不匹配,导致所构建的反演模型精度降低。为确定利用无人机多光谱影像反演水稻叶绿素含量的最优空间窗口,该研究分别采集水稻孕穗期、抽穗期和成熟期多光谱影像,以不同大小和形状的空间窗口对影像进行处理并计算多种植被指数,将不同窗口处理的植被指数与地面实测SPAD(soil and plant analyzer development)值进行相关性分析,将相关性最高的一组植被指数所对应的空间窗口确定为最优空间窗口,并以该组植被指数与地面实测SPAD值为依据,分别构建支持向量机、随机森林、极限学习机、广义线性模型和多元逐步回归模型,分析各模型在水稻各生育期对SPAD值的反演精度。结果表明:经过空间窗口处理后各植被指数与SPAD值间的相关系数与处理前相比均有较大提升,圆形空间窗口下各生育期的最优窗口半径分别为35、25、25个像素,方形空间窗口下各生育期的最优窗口边长分别为71、41、61个像素,方形窗口处理效果与圆形窗口近似;利用支持向量机模型反演水稻SPAD值的效果最优,且在孕穗期反演精度最高,决定系数为0.718,均方根误差为1.849,平均绝对误差为1.465。研究结果可为其他作物理化参数反演的空间窗口选择提供参考,为无人机利用多光谱监测作物长势、发展精准农业提供技术支持。展开更多
文摘When the remote sensing pixel is composed of multiple components and a non-isothermal surface, its directional signature of thermal-infrared radiation is mainly determined by the 3D structure of the pixel. In this paper, we present our simple directional thermal radiation model to describe the relation between the pixel thermal emission and the pixel’s component parameters, and invert the model to get the component temperatures. For the inversion algorithm, we focus on how to use the information of given observations in a more effective way. The information content in data space and parameter space is defined, and the transferring of information content in inversion procedure is studied. Our forward model and inversion method are validated using indoor directional measurement data.
文摘通过无人机多光谱影像反演农作物理化参数、动态监测作物长势是精准农业发展的重要方向。然而,由于无人机影像多具有较高的空间分辨率,地面采样点与影像上对应像素的空间范围往往不匹配,导致所构建的反演模型精度降低。为确定利用无人机多光谱影像反演水稻叶绿素含量的最优空间窗口,该研究分别采集水稻孕穗期、抽穗期和成熟期多光谱影像,以不同大小和形状的空间窗口对影像进行处理并计算多种植被指数,将不同窗口处理的植被指数与地面实测SPAD(soil and plant analyzer development)值进行相关性分析,将相关性最高的一组植被指数所对应的空间窗口确定为最优空间窗口,并以该组植被指数与地面实测SPAD值为依据,分别构建支持向量机、随机森林、极限学习机、广义线性模型和多元逐步回归模型,分析各模型在水稻各生育期对SPAD值的反演精度。结果表明:经过空间窗口处理后各植被指数与SPAD值间的相关系数与处理前相比均有较大提升,圆形空间窗口下各生育期的最优窗口半径分别为35、25、25个像素,方形空间窗口下各生育期的最优窗口边长分别为71、41、61个像素,方形窗口处理效果与圆形窗口近似;利用支持向量机模型反演水稻SPAD值的效果最优,且在孕穗期反演精度最高,决定系数为0.718,均方根误差为1.849,平均绝对误差为1.465。研究结果可为其他作物理化参数反演的空间窗口选择提供参考,为无人机利用多光谱监测作物长势、发展精准农业提供技术支持。
