Phase Unwrapping for Interferometric Synthetic Aperture Radar
We developed a phase-unwrapping algorithm applicable to interferometric synthetic aperture radar (InSAR) systems. Our approach has the advantage of removing noise while simultaneously unwrapping phases. Our algorithm has been incorporated into Mitsubishi Electric's latest InSAR system, which has been successfully tested in the field.
Background & Objective: Phase-unwrapping is a necessary step in obtaining a terrain height map from a satellite or airplane-based InSAR system. InSAR measurements provide a phase which is inherently ambiguous with respect to the addition or subtraction of units of 2*pi, and this ambiguity must be removed to recover a sensible height map. To further complicate the problem, InSAR phase measurements are also noisy. The phase-unwrapping problem can be attacked by finding the most probable set of "unwrapped" phases given the received phases, using a terrain model and a noise model. In our approach the terrain and noise models are explicitly defined by a "factor graph" and the most probable configuration is optimized iteratively using a combination of two algorithms, an iterated-conditional-modes (ICM) algorithm, and a minimum-cost-flow (MCF) algorithm. Our algorithm provably finds a locally optimal configuration.
Technical Discussion: Our phase-unwrapping algorithm is being used in the latest Mitsubishi Electric InSAR system, and is a key reason that the new system is able to generate much more accurate digital surface models. An experimental field observation was carried out in Tsukuba, Japan, and the new InSAR system incorporating MERL's phase-unwrapping algorithm achieved an accuracy of better than 50 cm at a spatial resolution of 30 cm. Detailed terrain features such as rice field terraces are accurately recovered.
Publications:
Technology Area: Imaging
Modification Date: August 13, 2008