Remote sensing over a wide area, such as a rainforest or arctic region, is often done by segmenting multispectral images from a spaceborne passive radar. In such applications, the number and types of segmented regions are often unknown, and boundaries between segmented regions are often fuzzy. An unsupervised segmentation approach utilizing a fuzzy clustering algorithm is presented in this paper, and applied to segmentation of images of arctic regions. The steps to improving processing speed, including preprocessing and reduction of feature space, are also presented. The experimental results with images from SSM/I sensor show that segmented regions found by this fuzzy clustering approach are stable and that the seasonal change can be monitored from the clustered results.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below