부산대학교 도서관

Hyperspectral dimensionality reduction is a significant pre-processing phase preceding complex data analysis. The present work is address, the design and implementation of new techniques for dimensionality reduction and segmentation on remotely sensed hyperspectral scenes. The enhanced clustering method Robust K-Means is performed in inter-intra clustering parts. The band reduction and the segmentation have performed this clustering method. The hyperspectral bands are clustered and a band that has uppermost variance from each cluster is single out. This process can reduce set of bands. Furthermore, RKM accomplished the segregation procedure on this reduced bands. Execution of this technique is assessed in various circumstances.