Unsupervised segmentation of multitemporal interferometric SAR images

This paper shows how to segment large data sets of multitemporal and interf erometric SAR images usi ng an unsupervised, fuzzy clustering method. An ad aptive feature extraction (principal component transformation) is employed which may drastically reduce the number of images and improves the final re sults. This also speeds up the fuzzy clustering iteration part considerably , The method is applied to data over two areas in Sweden: one typical urban area with forest and farmland surroundings and a forested area. The best c lassification accuracy is obtained when classifying the data into two class es, agreeing with the predictions of the cluster validity parameters used i n this study. The method always finds the dominating land-covers in the ima ges first. These are then subdivided as more clusters (classes) are identif ied, indicating that the segmentation is moderately hierarchical. The final classification results, between 65% and 75%, are comparable to those obtai ned in other studies. Analyzing the final cluster signatures reveals that t he current unsupervised method has several similarities with rule-based met hods.