Research Methods Of Remote Sensing Image Clustering Based On Non-symmetry Type-2 Fuzzy Sets

Remote sensing images are divided into satellite images and aerial images. Particularly, Remote sensing image clustering algorithm is a key task in remote sensing application technology systems. It is still difficult to meet the needs of practical application although it has been made rapid progress in theory and technology fields. In modern military, remote sensing images take an important place in the military images. We can get important military goals and hit them accurately on the ground in remote sensing images which can be gotten through satellite, airborne camera equipment, etc. As a result, high-tech in military revolution can improve the forces fighting capacity.Recently, fuzzy C-means clustering algorithm is a main remote sensing image clustering algorithm, which has been extensively investigated in image processing field and a series of related theory knowledge has been developed to enrich the gap in fuzzy c-means clustering algorithm. Essentially, the interval type-2 fuzzy C-means clustering algorithm is a local optimization technique. At the same time, selections of center positions and parameters for initial category have huge impact in the clustering results. However, the processing ability for asymmetric distribution needs to be improved, because of default symmetric for interval type-2 fuzzy membership function. Type-reduction processing is complex and has a low efficiency. In conclusion, several improvement methods have been put forward in this dissertation and applied in the segmemation of remote sensing image.(1) non-symmetry interval type-2 fuzzy sets has been proposed in this thesis.(2) The optimal selection of clustering number for non-symmetry interval type-2 fuzzy C-means clustering algorithm has been proposed to lower center uncertainty of type 2 fuzzy clustering algorithm.(3) Non-symmetry interval type-2 fuzzy C-means algorithm has been put forward to promote the performance and stability of clustering algorithmThe feasibility and superiority for non-symmetry interval type-2 fuzzy C-means clustering algorithm has been demonstrated by several experimental results.