Classification for Beijing-1 Micro-satellite's Multispectral Image Based on Semi-supervised Kernel FCM Algorithm

Most of remote sensing image data do not satisfy to Gauss distribution,and still the problems of the nonlinear,fuzziness and in lack of labeled data exist in the remote sensing image classification.A semi-supervised kernel fuzzy c-means(SSKFCM) algorithm is proposed to overcome these disadvantages.First,the SSKFCM algorithm is formed by involving semi-supervised learning technique and kernel method into the standard fuzzy c-means(FCM) algorithm.Then,IRIS data set and Beijing-1 micro-satellite's multispectral images are classified by those algorithms,such as k-Means(KM),maximum likelihood(ML),multiclass support vector machines(MSVM),semi-supervised support vector machines(S3VM),FCM,kernel FCM(KFCM),semi-supervised FCM(SSFCM) and SSKFCM.Finally,the classification results are estimated by corresponding indexes.The results indicate that the SSKFCM algorithm significantly improved the classification accuracy of remote sensing images compared with the others.