부산대학교 도서관

In this paper, the segmentation process is attentive on hyperspectral satellite images. The input hyperspectral image embodies of image data at peculiar frequencies over the electromagnetic spectrum. The existing tactics for satellite image segmentation include K-means, Fuzzy C-Means and Fast K-Means Weighted clustering algorithms are used. Besides, an integrated image segmentation process based on inter-band clustering and intra-band clustering has proposed. Furthermore, the inter band clustering is performed by above clustering algorithms, while the intra band clustering is executed using Particle Swarm Segmentation (PSO) clustering algorithm. Moreover, DB (Davies Bouldin Index) index is pertained to determine the number of clusters for the above clustering processes. The hyperspectral bands are clustered and a band (maximum variance) from each cluster is picked out. This forms the diminished set of bands. Finally, PSC performs the segmentation process on the diminished bands. Keywords: K-Means, Fuzzy C-Means, Fast KMeans Clustering and Particle Swarm optimization.
INTRODUCTION Hyperspectral image analysis has bloomed into one of the puissant and nimblest augmenting technologies in remote sensing domain for recent years. Hyper represent giant-sized measured wavelength band. Hyperspectral images [...]