A Novel Remote Sensing Image Classification Method Based on Semi-supervised Fuzzy C-Means

Because of the uncertainty of remote sensing image and ill-posedness for model, the traditional unsupervised classification algorithm is difficult to model accurately in the classification process. The pattern recognition methods based on fuzzy set theory can manage the fuzziness of data effectively, such as fuzzy c-means clustering algorithm. Among them, the type-2 fuzzy c-means algorithm has better ability to control uncertainty. However, the semi-supervised method can use prior knowledge to deal with ill-posedness of algorithms more suitable. This paper proposes the method based on the semi-supervised adaptive interval type-2 fuzzy c-means (SS-AIT2FCM). In the interval type-2 fuzzy algorithm, soft constraint supervision is performed by a small number of labeled samples, which optimizes the iterative process of the algorithm and mines the optimal expression of the data, and reduces the ill-posedness of the algorithm itself. Experimental results indicate that SS-AIT2FCM can get more accurate clusters and more clear boundaries in the remote sensing image of serious mixed pixels, have an effective results to suppress the phenomenon of “isomorphic spectrum”.