High-Resolution InSAR Building Layovers Detection and Exploitation

Abstract

Layover affects the quality of urban interferometric synthetic aperture radar (InSAR) digital elevation models. Moreover, it is generally difficult to interpret because of the superposition of several contributions in a single SAR pixel. In this paper, a novel technique for the extraction of building layovers is first presented. It makes use of the geocoding stage embedded in the InSAR processor. It is shown that building layovers create a regular pattern in the mapping counter, a map describing the number of occurrences of a SAR pixel in the elevation model. Its exploitation yields a generation of a layover map without the use of external supports. The integration in the processor with a limited additional computational load and the capability to isolate layover signatures are additional benefits. Layover patches are then individually analyzed toward a better understanding of the complex urban signal return. A spectral estimation framework is employed to assess the slopes superimposed in the patches. Fringe-frequency estimation is involved. A set of simulations made for a nonparametric (fast Fourier transform) and a parametric (multiple signal classification) technique is performed prior to testing on real data. It is demonstrated that in X-band, for a single interferogram, just one layover contributor, when it dominates over the others, can be extracted with a sufficient accuracy. The algorithms are tested on a TanDEM-X spotlight acquisition over Berlin (Germany).