Unsupervised Fuzzy C-Means Classification for the Determination of Dynamically Homogeneous Areas

Conventional image classification procedures are often inappropiate for the mapping of continuous phenomena like those appearing in remotely sensed images of the Earth surface. Geographical information, included that remotely sensed, is imprecise in nature. For the allowness of natural fuzziness of such an environment, a fuzzy sets algorithm may be used. This report deals with the development of a fuzzy unsupervised procedure which is applied to the classification of a set of NDVI time series images. The goal of this technique is to locate dynamically homogeneous areas. The results show that fuzzy classification can provide a more detailed information than classical "hard" classification. However, more attention is required at the interpretation stage, due to the continuous nature of the classes. We conclude that the method is most promising and worthy of consideration when a mapping of natural resources is necessary.