An open source Opticks plug-in for high resolution SAR imagery orthorectification and stereo measurements

Abstract

Nowadays, the availability of novel high-resolution synthetic aperture radar (SAR) spaceborne sensors offers new interesting potentialities for the acquisition of data useful for the generation of secondary products as digital surface models (DSMs), orthoimages, and displacement maps. SAR technology provides for low-cost, fast data acquisition and processing, independence from logistic difficulties, and night-and-day and all-weather functionality. These features are of crucial importance for the timely monitoring and management of disasters and emergencies such as geological, hydrological, and geophysical hazards. However, one of the most critical aspects in extracting useful and reliable information from SAR data is the image processing. Although several commercial software suites are available, in recent years the open source technology has confirmed a reliable and effective alternative for SAR processing and in general for geospatial information management. For instance, projects such as the Open Source Geospatial Foundation (OSGeo) offer to users, developers, and scientists powerful solutions based on costless, expandable, and customizable software. The goal of this work was to extend the capabilities of the Opticks remote sensing and imagery analysis software developing a plug-in (3DGeoCode plug-in) able to perform SAR precise image orthorectification (Orthotool) and to retrieve three-dimensional information starting from a stereo-pair (stereo measurement tool). At the moment of writing this article, the plug-in handles high-resolution SAR imagery acquired by TerraSAR-X and Radarsat-2 sensors. A complete description of the exploited algorithms is illustrated with special focus on showing the solutions adopted during the implementation. Moreover, a deep analysis of the accuracy achievable by the 3DGeoCode plug-in is reported showing the results obtained in several test sites. In particular, using a high-resolution lidar DSM and a TerraSAR-X spotlight stereopair over the area of Trento city (northern Italy), a root mean square error (RMSE) of approximately 2–3 m with respect to the reference DSM both for orthoimages and stereo measurements was achieved.