Refinement of interferometric SAR parameters using digital terrain model as an external reference
Refinement of interferometric SAR parameters using digital terrain model as an external reference
复制标题
DOI：
10.1016/j.isprsjprs.2021.02.017
复制
发表时间：
2021-05
影响因子：
12.7
通讯作者：
J. Uemoto
中科院分区：
文献类型：
--
作者：
J. Uemoto
合成孔径雷达 (SAR) 的交叉轨迹干涉测量 (CTI) 是一种有用的技术，已广泛用于检索数字表面模型 (DSM)。 CTI 观测的目的之一是测量目标场景中物体的高度，实现此目的的一种直接方法是从 CTI DSM 中减去数字地形模型 (DTM) 的外部参考。另一方面，在减法之前通常需要使用外部参考来细化 CTI 参数，即干涉 SAR (InSAR) 参数。为了避免因用于减法和细化的参考之间的地形高度的潜在差异而导致的不良降级，最好使用单个 DTM 作为公共参考。然而，如果在没有仔细考虑的情况下将 DTM 用作 GCP 的外部源，则无法正确细化 InSAR 参数，因为 CTI DSM 包含与 DTM 不同的对象。在此背景下，我们提出了一种基于 DTM 细化 InSAR 参数的新方法。该方法的特点是自动提取的垂直结构的足迹被选择作为地面控制点（GCP）候选者，并且通过用一种或两种正态分布拟合高度差的直方图来迭代选择预期位于地形水平的GCP。这使我们能够使用单个 CTI DSM 数据集和外部 DTM 来细化 InSAR 参数，而无需在此过程中进行任何交互操作。基于复杂场景的真实 CTI 数据集和激光 DTM 的实验证明了该方法的良好性能。测试场景地形区域高度误差的中值和四分位数范围分别为-0.18 m和1.81 m，这对于各种应用来说是可以接受的。发现在测试场景中使用两种新颖想法所带来的改进程度具有可比性。由于其固有的局限性，预计该方法对于城市和住宅区等发达地区是有效的。
Cross-track interferometry (CTI) for synthetic aperture radars (SARs) is a useful technique that has been widely used to retrieve a digital surface model (DSM). One aim of CTI observations is to measure the heights of objects in a target scene, and a straightforward approach for this purpose is to subtract an external reference of the digital terrain model (DTM) from the CTI DSM. On the other hand, the refinement of parameters for CTI, namely interferometric SAR (InSAR) parameters, using an external reference is often required before subtraction. To avoid undesired degradation caused by potential differences in the terrain height between the references used for subtraction and refinement, it is preferable to use a single DTM as a common reference. However, InSAR parameters cannot be refined correctly if a DTM is used as an external source for GCPs without careful considerations since a CTI DSM contains objects unlike a DTM. In this context, we propose a novel method that refines InSAR parameters based on a DTM. The features of the proposed method are that automatically extracted footprints of vertical structures are selected as ground control point (GCP) candidates and that GCPs expected to lie at the terrain level are iteratively selected by fitting the histogram of the height difference with one or two normal distributions. This enables us to refine InSAR parameters with a single CTI DSM dataset and an external DTM without any interactive operations in the process. An experiment based on a real CTI dataset for a complex scene and a laser DTM demonstrates the promising performance of the proposed method. The median and interquartile range of the height error in terrain areas for the test scene are −0.18 m and 1.81 m, respectively, which are acceptable for various applications. The degrees of improvement resulting from the use of two novel ideas in the test scene are found to be comparable. The proposed method is predicted to be effective for developed areas such as urban and residential areas owing to its inherent limitations.