Usability-Based SAR System Assessment: Framework and First Results

Abstract:
Human evaluation of SAR images as an indirect means for usability-based assessment of SAR systems from which the images arose, is time-consuming and costly. An assessment framework is presented which aims at finding a numerical evaluation to simulate, complement and partly replace the human evaluation of SAR images. In the human evaluation, expert SAR interpreters will be asked to solve a set of representative interpretation tasks on SAR images whose quality results from a specific SAR system and parameter setting. In the proposed assessment framework, it is planned to set up a dependency hypothesis by comparing a first quality metric derived from the interpreter evaluation and a second metric derived from the numerical evaluation. The SAR system component to be investigated first is the coding algorithm where the spatial and amplitude resolution are fundamental parameters. In this paper, we describe first results with respect to spatial resolutions necessary for reliably solving the initial set of representative interpretation tasks and their associated SAR test images. In future, these findings allow to perform subsequent evaluations with a smaller revised set of representative interpretation tasks and their associated images by avoiding those tasks that can always or never be solved.