A NEW CLUSTERING-ALGORITHM APPLICABLE TO MULTISPECTRAL AND POLARIMETRIC SAR IMAGES

We describe an application of a scale-space clustering algorithm to th e classification of a multispectral and polarimetric SAR image of an a gricultural site. After the initial polarimetric and radiometric calib ration and noise cancellation, we extracted a 12-dimensional feature v ector for each pixel from the scattering matrix. The clustering algori thm was able to partition a set of unlabeled feature vectors from 13 s elected sites, each site corresponding to a distinct crop, into 13 clu sters without any supervision. The cluster parameters were then used t o classify the whole image. The classification map is much less noisy and more accurate than those obtained by hierarchical rules. Starting with every point as a cluster, the algorithm works by melting the syst em to produce a tree of clusters in the scale space. It can cluster da ta in any multidimensional space and is insensitive to variability in cluster densities, sizes and ellipsoidal shapes. This algorithm, more powerful than existing ones, may be useful for remote sensing for land use.