Simulating images of passive sensors with finite field of view by coupling 3-D radiative transfer model and sensor perspective projection
Simulating images of passive sensors with finite field of view by coupling 3-D radiative transfer model and sensor perspective projection
复制标题
DOI：
10.1016/j.rse.2015.02.020
复制
发表时间：
2015-06
影响因子：
13.5
通讯作者：
T. Yin;N. Lauret;J. Gastellu-Etchegorry
中科院分区：
文献类型：
--
作者：
T. Yin;N. Lauret;J. Gastellu-Etchegorry
如果模拟数据在辐射测量和几何方面都不真实，则很难将实际遥感数据与模拟遥感数据进行比较。提出了一种考虑传感器视场内多方向采集的建模方法，以模拟有限视场下被动传感器获取的真实地球表面图像。在该方法中，DART(离散各向异性辐射传输)三维辐射传输模型(RTM)与三维透视投影相耦合。目前的RTM假设所研究的景观的所有部分都是沿着相同的方向观看的，尽管所有被动成像器都是在具有非零立体角度的视场中获取能量。此外，它们不能考虑相机模型及其图像投影几何(例如，相机的透视投影和交叉轨迹成像器的平行透视投影)。这种情况对于低传感器高度和大视场的机载捕获来说尤其成问题。我们的新建模方法解决了这个问题：进入传感器的光线可以来自不同的方向。为此，在光线跟踪过程中，每个被动传感器采集都被模拟为准确的观察方向，即从散射点到传感器位置的瞬时矢量。相机和跨轨道成像器都是为大多数经典配置建模的。通过这种实现，DART为不同的研究领域提供了原始的模拟和评估，包括：1.被动传感器成像；2.无人机(UAV)捕获的视频；3.遥感图像中的局部热点(HS)效应；4.模拟的正射校正透视投影图像之间的像素级比较；以及5.不同传感器高度的机载和星载系统获取的图像之间的辐射变化。这5点在这里得到了说明。对于任何配置，在模拟和获取的RS数据之间进行精确的像素级比较是该方法的另一个重要应用。
Comparison of actual and simulated remotely sensed data is difficult if simulated data are not realistic in terms of both radiometry and geometry. This paper presents a modeling approach that considers the multi-directional acquisition within sensor field of view (FOV) in order to simulate realistic images of Earth surfaces, as acquired by passive sensor with a finite FOV. In this approach, the DART (Discrete Anisotropic Radiative Transfer) 3-D radiative transfer model (RTM) is coupled with 3-D perspective projection. Current RTMs assume that all parts of the studied landscape are viewed along the same direction, although all passive imagers acquire energy in a FOV with a nonzero solid angle. In addition, they cannot account camera model and its image projection geometry (e.g., perspective projection for camera and parallel-perspective projection for cross-track imager). This situation is particularly problematic for airborne acquisition with low sensor altitude and wide FOV. Our new modeling approach solves this problem: rays that enter a sensor can come from various directions. For that, during ray tracking, each passive sensor acquisition is simulated for the exact view direction, which is the instant vector from the scattering point to the sensor position. Camera and cross-track imager are both modeled for most classical configurations. With this implementation, DART provides original simulations and assessments for various research domains, including: 1. Passive sensor imaging; 2. Video captured by unmanned aerial vehicle (UAV); 3. Local hot spot (HS) effect in a RS image; 4. Pixel-wise comparison between simulated orthorectified perspective-projection images; and 5. Radiance variation among images acquired by airborne and spaceborne systems with different sensor altitudes. These 5 points are illustrated here. Accurate pixel-wise comparison between simulated and acquired RS data, for any configuration is another important application of this approach.