خوشه بندي داده هاي فراطيفي با استفاده از الگوريتم بهينه سازي گروه گربه سانان

Clustering of hyperspectral data based on cat swarm optimization

The unique capabilities of hyperspectral images in expressing the properties of phenomena of earth surface, guides the researches of this branch toward developing methods that so soon as possible decreases the need of interference human factor in processing data. However clustering is one on, Hyperspectral images, Particle swarm of the most applicable methods in many of propounded processing in hyperspectral data such as classification and automatic recognition. Nevertheless, paying attention to high dimension of hyperspectral data, the traditional clustering methods have low efficiency and usually are trapped into local optima. The techniques of population based clustering because of random search, can overcome many problems of traditional clustering methods. One of the most novel techniques of swarm based optimisation is algorithm based on cat swarm. The Cat Swarm Optimization (CSO) algorithm is proposed by modeling two major behavioral traits of cats. Existence of two modes in CSO, namely tracing and seeking modes not only encourage the capability of global search, but promote the ability of local search. Simultaneously global and local search have caused, this algorithm have better performance in comparison with most of co-level optimization algorithms such as particle swarm optimization. In this paper a clustering method is presented based on cat swarm optimization in order to clustering of hyperspectral data. Experimental results on hyperspectral data illustrate clustering based on CSO have distinctly better performance than K-means and so it have partly better performance than clustering based on PSO.