Two-stage sequence classification of PolInSAR imagery

In this paper, we present a two-stage scheme for supervised classification of polarimetric interferometric synthetic aperture radar (PolInSAR) imagery. In the first stage, a regularized logistic regression classifier is employed to generate probability vectors of object labels with polarimetric and interferometric features, respectively. The soft outputs (probability map) of previous logistic classifier with different features are concatenated as the input features of the second stage classifier-SVM classifier, which provides the final classification. We compare the two-stage methods against the baseline method and show its effectiveness.