Direct Geo-Rectification of Regional-Scale SAR Mosaics Using SRTM DEM Data

SAR mosaics are often not precisely geo-referenced because topographic distortions were not removed during the mosaicking process due to the lack of adequate digital elevation models (DEMs). The Shuttle Radar Topography Mission (SRTM) has recently provided high-resolution DEM data with nearly global coverage, and makes it possible to geo-rectify SAR mosaics. Though techniques are available for rectifying individual scenes of SAR imagery using DEM data, these methods encounter difficulties when rectifying SAR mosaics because abrupt geometric discontinuities occur in SAR mosaics at scene boundaries. This paper introduces an automated direct method to directly removing geometric and topographic distortions and discontinuities from SAR mosaics and producing ortho-rectified mosaics, without accessing original SAR images. The proposed method was applied to geo-rectifying the Amazon basin GRFM (Global Rain Forest Mapping project) SAR mosaic produced from thousands of JERS-1 SAR images. Validation results show that 1-pixel (i.e., 92 m) positioning accuracy (root mean square error) was achieved in both the high Andes Mountains and the flat Atlantic estuaries, compared to the 16-pixel error (i.e., 1,380 m) of the original mosaics. The proposed procedures were implemented at a high level of automation with little human intervention. The large data volume represented by the Amazon basin mosaic is manageable on a PC computer. With the release of high-resolution SRTM DEM data, the proposed method is readily applicable to SAR mosaics located between 56S to 60N.