Forest and non-forest discrimination using PolSAR data based on K-Wishart distribution

Based on the circular Gaussian assumption, the polarimetric covariance matrix of polarimetric synthetic aperture radar (PolSAR) is found having a complex Wishart distribution. It fits well for measurements over homogeneous regions, but often fails over heterogeneous backscattering media by SAR especially with a finer resolution. Accordingly, inadequacy of the Wishart model to match the data has been found in forest areas. In this case, the complex K-Wishart distribution has shown its particular attractiveness. It gives a more accurate and concise function to estimate the statistics of the full covariance matrix. Based on the expectation maximization (EM) algorithm, this study uses the complex K-Wishart probability density function (PDF) to discriminate the forest and non-forest. The effectiveness of the proposed method has been demonstrated.