PARAMETRIC AND NON-PARAMETRIC CLASSIFICATION METHODS FOR FULL POLARIMETRIC SAR IMAGES

Timely land use and land cover (LULC) information is essential for urban planning and management and it is important for local governments to create policies that will enable the maintenance of good balance between land development and environmental protection. PolSAR measurements achieve better classification results than single polarization SAR. The classification of PolSAR images has become an important research topic since PolSAR images have been made available through ENVISAT ASAR, ALOS PALSAR, and RADARSAT-2. However, most of the existing orbital SAR systems are single-frequency types and may create confusion during the separation and mapping of LULC classes; this confusion stems from the limited information obtained by single-frequency systems.To overcome the difficulty presented by single-frequency SARdata, some researchers utilized polarimetric SAR (PolSAR) data to investigate LULC informationThis paper provide three different classification methods provided by RADARSAT-2 PolSAR data. The two parametric classification including Freeman and Durden and the Wishart unsupervised classification and also a non-parametric neural network clustering. The results show that unsupervised wishart classification shows an important improvement in the description of the different types of natural media encountered in a forest scene, While the Unsupervised Freeman-Durden Wishart Classification has shown more details in the urban area.