Neighbor sample membership weighted KFCM algorithm for remote sensing image classification

Clustering accuracy of the Kernel Fuzzy C-Means (KFCM) algorithm is affected by its equal partition trend for data sets. A Neighbor Sample Membership Weighted KFCM (NSM-WKFCM) algorithm is achieved by introducing the weighted information of the neighbor sample membership into the standard KFCM algorithm in this paper. A set of Beijing-1 micro-satellite´s multispectral images is adopted to be classified by the KFCM and NSM-WKFCM algorithms. Experimental results indicate that the NSM- WKFCM algorithm significantly improve the unsupervised classification ability of remote sensing images compared with the KFCM algorithm.