摘要:
采用干涉相位直方图（phase histogram，PH）技术，原则上仅需单极化、单基线（或少量干涉图）即可获取低分辨率的森林垂直结构，然而，为了应对由机载轨道不稳定而导致垂直波数剧烈变化的问题，提出基线优选的策略，通过约束模糊高筛选合适的干涉基线，重建全覆盖实验区的相位直方图结果，并基于欧洲空间局TomoSense机载全极化合成孔径雷达（synthetic aperture radar，SAR）数据集，研究验证所提方法在长波机载SAR数据下获取森林3D垂直结构与森林高度的可行性。实验结果表明，PH技术在合适的干涉基线条件下，能够获取低分辨率的可表征主导散射体特征的3D后向散射能量剖面（即森林垂直结构）。同时也能够获取一定精度的森林高度产品，以机载激光雷达森林高度为参考，P波段与L波段数据估计的森林高度均方根误差分别为4.6 m和5.2 m。研究表明，PH技术能够通过少量基线数据获取森林垂直结构，具备未来星载高分辨率SAR卫星森林监测任务中广域森林制图的潜力。
Abstract:
Objectives: The primary tasks include: (1) Proposing multi-baseline optimized phase histogram (PH) technique to explore the 3D imaging capabilities of low-frequency airborne synthetic aperture radar （SAR） data for forested areas and to estimate canopy height model (CHM); (2) using a simplified physical model to simulate and explain the principles and limitations of the PH technique. Methods: Introducing a novel technique, the PH method, for estimating the 3D vertical structure and canopy height of forested areas.The PH technique leverages the phase-to-height relationship derived from interferometric phase and vertical interferometric wavenumber to assign each pixel to a specific horizontal height layer. By accumulating the magnitude of all pixels within a given spatial window at the same height layer, it approximates the 3D backscatter profile of the forest. To address the limitations of single-baseline observations and the significant variation in vertical wavenumber caused by unstable airborne platforms, a full-baseline-based optimized multi-baseline interferometric combination strategy is proposed. This strategy restricts the height am biguity range to obtain phase histograms that completely cover the experimental area. Results: The results indicate that, under appropriate interferometric baseline length, the PH technique can retrieve an approximate coarse-resolution vertical structure of the forests, with the 3D backscatter power profiles characterizing the dominant scatterers. Additionally, the proposed method can provide quite good forest height estimates. Specifically, taking LiDAR forest height as the reference value, the average root mean square error of estimated forest height using P-band and L-band data are 4.6 m and 5.2 m, respectively. Conclusions: Although the PH technique cannot precisely separate ground and canopy signal, it can still characterize, to some extent, the distribution of scatterer density and scattering energy across horizontal height layers within a spatial window. This relationship holds potential for further exploration in forest height inversion and biomass estimation. Overall, as a novel method for retrieving forest vertical structure with a limited number of interferograms, the PH technique could serve as a viable alternative in future spaceborne SAR satellite missions for forest monitoring.
百度学术
下载:
