A radargrammetric orientation model for SAR high resolution imagery

Digital Surface and Terrain Models (DSMs/DTMs) have large relevance in some territorial applications, such as topographic mapping, spatial and temporal change detection, feature extraction and data visualization. DSMs/DTMs extraction from satellite stereo pair offers some advantages, among which low cost, speed of data acquisition and processing, surveys of critical areas, easy monitoring of wide areas, availability of several commercial software and algorithms for data processing. In particular, the DSMs generation from SAR data offers the significant advantage of possible data acquisition during the night and in presence of clouds. The availability of new high resolution SAR spaceborne sensors offers new interesting potentialities for the acquisition of data useful for the generation of DSMs. Two different approaches may be used to generate DSMs from SAR data: the interferometric and the radargrammetric one, both using a couple of images of the same area acquired from two different points of view. The aim of this work is the development and the implementation of a rigorous radargrammetric model for the orientation of SAR imagery, suited to the Digital Surface Model generation. The model performs a 3D orientation based on two range and two zero-Doppler equations; the results presented are related to the orientation of SAR stereo pairs in slant range and zero-Doppler projection, acquired in SpotLight mode (1 m ground resolution). The model has been implemented in SISAR (Software per Immagini Satellitari ad Alta Risoluzione), a scientific software developed at Geodesy and Geomatic Institute of the University of Rome La Sapienza. This software was at first devoted to the orientation of high resolution optical imagery, and in the last year it has been extended also to SAR imagery. Moreover a tool for the SAR Rational Polynomial Coefficients (RPCs) generation has been implemented in SISAR software, similarly to the one already developed for the optical sensors. The possibility to generate RPCs starting from a rigorous model sounds of particular interest since, at present, the most part of SAR imagery is not supplied with RPCs, although the Rational Polynomial Functions (RPFs) model is available in several commercial software. The RPCs can be an useful tool in place of the rigorous model in processes as the image orthorectification/geocoding or as the DSMs generation, since the RPFs model is very simple and fast to be applied. The model implemented has been tested on COSMO-SkyMed and on TerraSAR-X images; both satellites are able to acquire imagery in SpotLight mode. The accuracy evaluation shows that the mean vertical accuracy is at 3.0 - 3.5 m; this preliminary results are satisfying, considering the mean accuracy of the available GPs. An investigation using more accurate GPs, for example GPS points, is recommended, although it has to be underlined that also the uncertainty of the GPs collimation on the images (1-3 pixels) could affect the results. As regards the model performance varying the independent sets of GCPs, the software shows a stable behavior and the increase of GCPs number does not improve the results remarkably. The commercial software OrthoEngine provides basically the same level of accuracy of SISAR, whereas a significant difference between the two software is that OrthoEngine needs 8 GCPs at least for the orientation of SAR images, whereas SISAR gets good results also using few points (3 or 5 GCPs). The application of RPCs model to SAR stereo pairs gives good results, absolutely comparable with those derived through the radargrammetric rigorous model, what proves the effectiveness of the RPCs generation tool implemented in SISAR.