Optimization of KFCM Clustering of Hyperspectral Data by Particle Swarm Optimization Algorithm

Hyperspectral sensors, by accurate sampling of object reflectance into numerous narrow spectral bands, can provide valuable information to identify different landcover classes. nevertheless, classification of these data has some problems. in particular, one of the most well-known of them is not having adequate training data for learning of classifiers. one possible solution to this problem is the use of unsupervised classification such as kernel based fuzzy c-means (kfcm). kfcm is a kernelized version of fcm algorithm, which usually, has better performance. however, in case of hyperspectral data, accuracy of the kfcm decreases because of high dimensionality of data and its kernel parameter. in this paper, the objective is to use the kfcm clustering and optimize it based on data dimensionality and kernel parameter. to optimize this algorithm with respect to the kernel parameter and data dimensionality, particle swarm optimization method (pso) is introduced. in other words, pso is a powerful optimization tool inspired from bird’s behavior, which can find global optimum. in this study, two new methods are defined to optimize kfcm with respect to kernel parameter and data dimensionality. the results show that the proposed methods have a better performance than the kfcm.