Classification of Landsat remote sensing images by a fuzzy unsupervised clustering algorithm

摘要: Abstract The classification of each pixel in a Landsat image to one the land cover types by conventional clustering techniques is highly inappropriate due low resolution images and multiplicity terrain. concept fuzzy logic provides flexible solution this problem. This paper presents new two-pass unsupervised algorithm incorporated theory. In first pass mean vectors different representing their geographic attributes are derived. second membership grade belonging computed based on distance between its gray-value vector type. Experimental results show that developed produces more reasonable phenomenon interpretation than traditional hard partition techniques.