Integration of COSMO-SkyMed and GeoEye-1 Data With Object-Based Image Analysis

This paper describes the potentialities of data integration of high spatial resolution multispectral and single-polarization X-band radar for Object Based Image Analysis (OBIA) using already available algorithms and techniques. GeoEye-1 multispectral images (0.5/2.0 m) and COSMO-SkyMed stripmap images (3.0 m) were collected over a complex test site in the Venetian Lagoon, made up of an intricate mixture of settlements, cultivations, channels, roads, and marshes. The validation confirmed that the integration of optical and radar data substantially increased the thematic accuracy (about 20-30% for overall accuracy and about 25-35% for k coefficient) of multispectral data and, unlike the outcomes of some new researches, also confirmed that with appropriate pre-processing traditional OBIA could be applied also to X-band radar data without the need of developing ad hoc algorithms.