Robustness of wavelet-based stereo matching for variable acquisition geometries using simulated SAR images

It has been shown that digital surface models (DSMs) generated using stereo SAR can be used to ease the phase-unwrapping process during interferometric height model generation. In addition, wider availability of combined stereoscopic and interferometric coverage generated by air- and spaceborne sensors makes the combined use of the two techniques more feasible. This paper describes a series of experiments whereby different stereo acquisition geometries were simulated for an airborne X-band SAR sensor. The product of such a simulation, taking the topography, flight geometry, and local illuminated area into account, is a pair of amplitude images in slant range geometry. The use of such simulations allows rigorous control of error sources. A phase-based, multi-resolution image-matching technique built on the discrete wavelet transform is used to measure the parallax field between simulated image pairs, and a geocoded height map is generated in each case. The robustness of the matching algorithm across variable sensor configurations is thereby determined. Same-side stereo geometry is assumed, as is a typical airborne flight height of 2.8 km. Matching accuracy and the subsequent height accuracy of the DSMs are evaluated by varying the sensor incidence angles and stereo intersection angle. The optimal flight geometry for a test area in Switzerland is determined.