Geometric Rectification And Geocoding Of Airborne Polarimetric Sar Data. The Maestrol Case

Abstract

Problem definition : A strict parametric mathematical formulation of the SARSLAR geometry is possible if both the parameters of interior orientation, which relate image coordinates and object coordinates, and the parameters of exterior orientation, which relate sensor and object coordinates, exist Airborne SAR data are only seldom accompanied by sufficient auxiliary data to allow the application of the paramemc model. The problem becomes even more complicated if not all object related parameters are present. This paper describes primarily the effort to Georeference the airhe polarimetric SAR data collected by the JPL-AIRSAR system during the MAESTRO-1 experiment in the summer of 1989 over the German Freiburg testsite. In the case one wants to generate SAR datasets which can be overlaid on existing map based thematic data sets one has to Geocode the SAR data. The geocoding process involves the resampling of the polarimetric SAR data. Weay to evaluate the effect of several resampling algorithms on a number of polarimetric quantities which are commonly derived from polarimetric SAR data in application studies. Methodology : Georectification - The auxiliary data provided with AIRSAR data does not provide sufficient parameters (navigational, attitude, geo-positioning) to describe parametrically the exterior orientation. Therefore, a hybrid method is appliq based on a number of supplied navigational parameters and minimal one groundcontrol point which is referenced in both image and map space. The construction of the exterior orientation over the Freiburg testsite required the use of a digital elevation model. The calculation of the intersection of the SAR signal wavefront and the digital elevation model allowed, within certain errorbounds, to reconstruct the parameters of the exterior orientation. Geocoding - Emphasis was placed on quantitatively evaluating the resampling effect on the polarimetric signatures of selected targets. A number of different resampling schemes have been applied, Nearest Neighbor, Bi-cubic Spline, Bi-linear and Dynamic Resampling. Care has been taken to keep the resampling frequency minimal Nyquist, Results: Georectification - The georectification procedure generates per slant range pixel the corresponding map coordinates, elevation above datum, local incidence angle and ground'range projected pixel dimensions. A geometric error budget is presented. Geocoding - The effect of different resampling approaches, on the polarimetric signatures and derived quantities over a number of areas is presented. It is shown that in order to maintain the full information potential of polarimetric SAR data this resampling step has to be carried out on the cross products of the elements of the scattering matrix, or on the derived Stokes matrix representation.