Using Multiple View Angles to Improve the Accuracy of UAVSAR Land Cover Classification

Inhalt:
The NASA Jet Propulsion Lab UAVSAR has been acquiring full-polarimetric L-band SAR imagery since its first mission in 2008. The remarkable features of this radar suggest the use of its imagery for PolSAR land cover classification. In this paper, we explore UAVSAR data for a study area in Hawaii. The high variability of local incidence angle caused by the airborne SAR geometry and extreme topography of the region must be addressed before standard classification techniques can be applied. Multiple images from different view angles are compared to study the effects of incidence angle on the polarimetric properties. Finally, the multiple, orthogonal view angles are synthesized into a composite image that yields improved land cover classification.