High-resolution Spaceborne SAR Interferometry For DEM Generation And Updating

Digital Elevation Models (DEM) are used as basic data in Geomatics. DEMs play an important role in scientific research, economic construction and security applications. With the construction and operation of high-resolution SAR (Synthetic Aperture Radar) satellites, like ALOS/PALSAR, COSMO-SkyMed, TerraSAR-X/TanDEM-X, spaceborne InSAR (Interferometric SAR) becomes a powerful technology for high-resolution and high accuracy DEM mapping on a global scale.Although the spatial resolution, the interferometric coherence and the orbit accuracy have been greatly improved with the high-resolution spaceborne SAR, there are still unresolved problems due to the observation mode and the side-looking geometry of spaceborne SAR:1) in repeat-pass interferometry, the atmospheric effect will introduce significant elevation error;2) the length of the perpendicular baseline contradict the reliability of the phase unwrapping and the accuracy of the elevation measurement;3) the foreshortening, layover and shadow casting caused by SAR side-looking geometry will result in voids in the InSAR DEM.In order to overcome these three problems above and to promote the topographic application of high-resolution spaceborne InSAR in China, it is necessary to develop new methods and techniques to improve the InSAR DEM accuracy and to meet the surveying and mapping industry standard in China as well as the corresponding international standards. Based on the InSAR theory and principles, we analyze the advantages of high-resolution spaceborne SAR for interferometry, carry on a thorough research in the following three aspects and propose methods and techniques to solve the problems mentioned above.First, for repeat-pass InSAR atmospheric correction, we analyze the characteristics of tropospheric and ionospheric delay of SAR signals. Based on the spatial distribution and statistical characteristic of the atmospheric phase screen (APS), we propose an approach containing a phase/elevation regression and a combined spectral filter to estimate the APS on a single high-resolution interferogram. The APS is distinguished into two parts-stratification and turbulence, which are estimated by the regression and the filter respectively and are then removed. The proposed approach only needs a low-resolution DEM as external data, and can effectively improve the InSAR DEM accuracy without meteorological observation data or models. Therefore, the proposed approach has a wide applicability. Second, for the contradiction between the length of the perpendicular baseline, the reliability of the phase unwrapping and the accuracy of the elevation measurement, we study the multi-baseline InSAR method for DEM generation. Based on the existing multi-baseline methods and the probability distribution of the interferometric phase, we propose a maximum a posteriori (MAP) estimation method and define the prior probability distribution of elevation from a low-resolution external DEM. The prior probability distribution is used to improve the multi-baseline InSAR accuracy and reliability, and the MAP estimation avoids phase unwrapping. In order to adapt the MAP method to high-resolution spaceborne SAR data, we present a strategy for multi-baseline image pair combination and coregistration. The elevation/phase rational polynomial model, the elevation likelihood probability look-up table and the variable step size searching to speed up the MAP estimation.Third, for the elimination of InSAR DEM voids caused by foreshortening, layover, and shadow, we study cross-heading (ascending/descending) InSAR methods for DEM generation and discuss two approaches:the joint estimation and the DEM fusion. Since the spaceborne SAR observations are lacking ground control points (GCP) in most cases, we propose an approach for cross-heading orbit InSAR DEM fusion without GCP. The approach uses a low-resolution external DEM as reference to correct the geolocation error between the ascending and descending InSAR pair and to achieve the coregistration of the DEMs acquired from cross-heading orbit. Meanwhile, we derive the weight for the maximum likelihood fusion of cross-heading orbit InSAR DEM, which will improve the DEM accuracy in the process of eliminating the DEM voids.Our research results improve the methods and techniques for InSAR DEM generation, and build a systematic processing flow for high-resolution spaceborne InSAR based on the above mentioned studies. The proposed methods are verified through the InSAR DEM generation experiments on ALOS/PALSAR, COSMO-SkyMed and TerraSAR-X high-resolution data. The experiments also prove the application level of high-resolution spaceborne InSAR in topographic mapping:the ALOS/PALSAR multi-baseline InSAR DEM can meet the China’s1:50000DEM standard and the US DTED-2standard, and the COSMO-SkyMed/TerraSAR-X DEM can meet the China’s1:25000DEM standard and the US HRTI-3standard.