SAR Tomography With Small Data Stack by Refining the Reference Network

This letter introduces a reliable 3-D reconstruction for buildings using a small data stack in synthetic aperture radar tomography (TomoSAR) based on a revised reference network (RN) method. The RN method can realize atmospheric phase error correction by persistent scatterer (PS) phase differencing. However, when dealing with a limited number of images in a small data stack, the inaccurate inversion of relative elevation for local PS arcs can result in error propagation during network integration, leading to compromised tomography results. To address this issue, we have refined the traditional RN method by introducing a new geometric constraint between PSs. In the proposed method, we use the relative elevation estimated from PS arc differential signal to accurately determine the spatial distance between PSs. PS arcs that exceed the geometric constraint are eliminated to assist in adjusting the RN. The effectiveness of this optimization method is demonstrated through TomoSAR experiments conducted on a high-resolution TerraSAR-X SAR data stack in Shenzhen, China. Consequently, the method presented in this letter is highly suitable for processing small data stacks compared with the previous RN method.