Comparison of Algorithms and Input Vectors for Sea-Ice Classification with L-band PolSAR Data

Convolutional neural network (CNN) and fuzzy c-means (FCM) are applied to L-band PolSAR data in advanced-melt and winter phases, respectively, to classify sea-ice type. Several types of input vector are derived from the covariance matrix to improve the accuracy rate of classification. The parameters in algorithms and input vectors are also analyzed.