Formation of fuzzy land cover objects from TM images

Spatial features are usually simplified as crisp objects in most GIS models. However, this kind of abstraction is not accurate enough, since in many cases natural phenomena are continuous and there are no determinate boundaries between different spatial objects. The handling of these fuzzy spatial objects is one of the most important questions in GIS. The first step in modeling fuzzy spatial objects is to identify and obtain these objects. This paper discusses the general procedure for forming fuzzy spatial objects and proposes a composite method to generate fuzzy land cover from TM images. The method is developed based on a series of image analyses including supervised classification, fuzzy convolution and rule-based processing. It is verified in a test area of Sanya city of China. The result shows that the methodology is applicable for forming fuzzy land cover objects.