Classification from Airborne Sar Data Enhanced by Optical Image Using an Object-oriented Approach

Classification from Synthetic aperture radar(SAR) is known to be difficult because of the speckle and variation in scattering coefficient with incidence angle, especially for high resolution airborne SAR. In western China, most areas are cloud cover perennially, which hinders acquisition of optical images. Airborne SAR data will be a good choice for big scale mapping purpose. To make effective use of airborne SAR in the area with cloud cover, it is essential to find a proper way to classify SAR imagery accurately. This paper present a general airborne SAR classification frame, which is proved effectively by the following test. This frame consists in two main aspects: 1) Information enhancement, including SAR speckle removing; single polarisation airborne SAR merged with optical multispectral data. The improved spectral information was contributed to the classification afterward. 2) Object-oriented classification, the enhanced image is analysed by object-oriented approach of eCognition, It is based on fuzzy logic, allows the integration of different object features, such as spectral values, shape and texture.