Segmentation of IKONOS Images by Genetically Derived Hard and Fuzzy C-Means Clustering Algorithm

N.N. Kachouie and J. Alirezaie (Canada)


Genetic Algorithm, Fuzzy c-means, Hard c-means,Clustering, Segmentation, IKONOS


A genetic segmentation algorithm is implemented, which is a hybrid technique. This paper describes a clustering method that utilizes hard and fuzzy clustering algorithms. The performance of the algorithm is optimized using genetic algorithm which searches the best cluster centers to initialize the fuzzy and crisp partition matrices in place of random initialization. The proposed algorithm is almost insensitive to initial conditions and performs better than the formers. The proposed approach provides accurate clustering results for gray-level and color images. The algorithm is tested on the several subsection images of a Georeferenced IKONOS image using different urban scenes in the area of the International Pearson Airport in Toronto, Canada. Proposed method partitions different land cover regions and provides proper classification of pixels that is an important step in the realm of satellite imagery. The comparison between segmentation results of Fuzzy c-means Genetic Algorithm (FGA) and Fuzzy C Means (FCM) is presented. Hard c-means Genetic Algorithm (HGA) and Hard C-Means (HCM) are compared as well.

Important Links:

Go Back