Clustering synthetic aperture radar (SAR) imagery using an automatic approach

Synthetic aperture radar (SAR) imagery has been shown to be useful for land surface applications. Similar to optical imagery, unsupervised or supervised algorithms can also be used to classify SAR data. Supervised classification methods require a priori information, which is usually not available, especially over a large area. Similarly, existing unsupervised classification methods based on clustering algorithms (e.g., K-means and ISODATA) require control input parameters, such as the number of clusters, which are difficult to obtain over a large area. In this paper, we present a new automated clustering method for SAR imagery that does not require input of control parameters. The main advantage of this method over other methods is the capability to automatically determine the number of statistically separable clusters in a SAR image. The performance of the method is assessed over two test sites.