부산대학교 도서관

The proposed work addresses a novelty in techniques for segmentation of remotely sensed hyper-spectral scenes. Incorporated inter band cluster and intra band cluster techniques has investigated. With a new constrain validate the new segmentation methods in this proposed work, the fast K-Means is used in inter clustering part. The inter band clustering is carried out by fast K-Means methods includes weighted and careful seeding procedures. The intra band clustering processed using Particle Swarm Clustering algorithm with enhanced estimation of centroid. Davies Bouldin index is used to determine the number of clusters in the mentioned clustering strategies. The hyper-spectral bands are clustered in order to reduce the band size. In next phase, the above said enhanced algorithm carried out the segmentation process in the reduced bands. In addition, statistical analysis is carried out in various scenarios. Keywords: Fast K-Means, Weighted K-Mean, Careful seeding, Particle swarm clustering Davies Bouldin index
1. Introduction Hyper-Spectral Imaging (HSI) is a new remote sensing scheme that develops hundreds of images, tantamount to divergent wavelength channels, in a similar region on the surface of the [...]