Unsupervised classification of polarimetric SAR image with dynamic clustering: An image processing approach

This paper proposes a new unsupervised classification approach for automatic analysis of polarimetric synthetic aperture radar (SAR) image. Classification of the information in multi-dimensional polarimetric SAR data space by dynamic clustering is addressed as an optimization problem and two recently proposed techniques based on particle swarm optimization (PSO) are applied to find optimal (number of) clusters in a given input data space, distance metric and a proper validity index function. The first technique, so-called multi-dimensional (MD) PSO, re-forms the native structure of swarm particles in such a way that they can make inter-dimensional passes with a dedicated dimensional PSO process. Therefore, in a multi-dimensional search space where the optimum dimension is unknown, swarm particles can seek both positional and dimensional optima. Nevertheless, MD PSO is still susceptible to premature convergence due to lack of divergence. To address this problem, fractional global best formation (FGBF) technique is then presented, which basically collects all promising dimensional components and fractionally creates an artificial global-best particle (aGB) that has the potential to be a better “guide” than the PSO’s native gbest particle. In this study, the proposed dynamic clustering process based on MD-PSO and FGBF techniques is applied to automatically classify the color-coded representations of the polarimetric SAR information (i.e. the type of scattering, backscattering power) extracted by means of the Pauli or the Cloude–Pottier decomposition algorithms. The performance of the proposed method is evaluated based on fully polarimetric SAR data of the San Francisco Bay acquired by the NASA/Jet Propulsion Laboratory Airborne SAR (AIRSAR) at L-band. The proposed unsupervised technique determines the number of classes within polarimetric SAR image for optimal classification performance while preserving spatial resolution and textural information in the classified results. Additionally, it is possible to further apply the proposed dynamic clustering technique to higher dimensional (N-D) feature spaces of fully polarimetric SAR data.