Applying maximum entropy methods to aid in classification of polarimetric SAR imagery

This research presents a technique intended for use in classification of remotely sensed SAR data. As an input for the technique, the user provides a set of polarimetric radar data of a particular distributed target. The output produced is an analytical, multidimensional probability density function (pdf). To find the pdf that generated a particular set of data, maximum entropy techniques are used. These techniques work by requiring that in addition to constraints provided by the data, the authors maximize the entropy of the probability density function. The authors apply this technique to SAR data, with the intent of eventually using this characterization in the training phase of a supervised classifier