Maximum Likelihood Classification of Single High-resolution Polarimetric SAR Images in Urban Areas

In this work, our aim is to assess the potential of a single polarimetric radar image of high spatial resolution for the classification of urban areas. For that purpose, we concentrate on a fine definition of urban land cover types including ground classes corresponding to different roof types and we test several supervised classification algorithms. In particular, we deal with maximum likelihood classification using several polarimetric and textural indices. At first, we propose a state-of-the-art statistical analysis of polarimetric synthetic aperture radar (SAR) data to study the statistical behaviours of these indices. We consider the Gauss, log-normal, Beta 1, Weibull, Gamma, K, and Fisher statistical models and estimate their parameters using two methods: maximum likelihood estimation (MLE), and method of log-moment (MoLM). The Fisher probability density function (pdf) is able to properly model all the descriptors. Then, we propose to introduce this information in an adapted supervised classification scheme based on maximum likelihood and the Fisher pdf. We compare the classification results with the Wishart-based maximum likelihood algorithm, a Gaussian-based one and SVM (support vector machine). Our experiments are based on an image of a suburban area, acquired by the airborne RAMSES SAR sensor of ONERA, the French Aerospace Lab. The results highlight the potential of such data to discriminate urban land cover types, and the overall accuracy reaches 84%. However, the results from the tested classification methods show a problematic confusion between roofs and trees. Some possible solutions are discussed at the end of this paper.