From complexity to clarity: visualizing PolInSAR data for environmental monitoring

Abstract

This paper addresses the challenge of visually interpreting Polarimetric Interferometric Synthetic Aperture Radar (PolInSAR) data for environmental monitoring. PolInSAR data provides valuable information on geophysical properties such as soil moisture, surface roughness, and vegetation height. While automated techniques can be used for land cover classification, human visual interpretation remains crucial for considering contextual information and integrating domain knowledge in data exploration. However, visual interpretation of PolInSAR data is challenging as a single image representation captures only a fraction of the information. To address this, we propose combining polarimetric and interferometric feature extraction and dimension reduction techniques. By projecting multi-dimensional feature representations into a 3-dimensional feature space using Uniform Manifold Approximation and Projection (UMAP), followed by automatic color mapping in CIELCh color space, comprehensive image representations are generated that facilitate land cover identification. Applied to multifrequency PolInSAR data acquired over the East-Frisian island Baltrum, our approach enables easy recognition of various types of salt marshes, dune vegetation, and tidal flats.