Classification of spectropolarimetric imagery based on fuzzy cluster and evidence theory

An unsupervised classifying algorithm is proposed using spectropolarimetric information. The algorithm can be described as follow: first, three parameters are clustered by fuzzy C-mean clustering (FCM), which include the intensity of radiation, the degree of linear polarization and the phrase of linear polarization. Second, basic belief assignment functions are constructed based on the cluster result and polarize characters of object's surface. Finally, a fusion algorithm is implemented using weighted distribution method. The results of experiment verify the effectiveness of the algorithm.