SEGMENTATION ANALYSIS OF PRINCIPAL COMPONENT WITHTRANSFORMATION KARHUNEN-LOEVE FOR EFFICIENCY OF PRESENTATION AND CLASSIFICATION OF IMAGE OF REMOTE SENSING. This paper proposes efficiency ofpresentation and classification of image of remote sensing by conducting segmentation andtransformation of principal component. Conditions desired are first doing partition of all bands tobecome a number of subgroups which have the high correlation; second, conducting transformation ofeach subgroup, one band dissociated to be made by a guide get choice characteristic. Then distinguishchoice of transformation return to be accepted as by a comparison reduce appropriate data and yield threemost component of significant for the colour presentation. Scheme of addition of segmentation inreducing calculation for election of compared to by done to be better characteristic of Conventional TKU.Reducing a number of characteristic quicken process of classify and process will not experience oftrouble to all data of hyperspektral of is although data of training sample limited. Result obtainedinterconnected of work of terminology of classify is accuration, speed and quality of presentation drawbased on data Compact Airborne Spectographic Imager (CASI).Keywords: feature selection, hyperspektral, image classification, principal component analysis, analisis Karhunen-LoeveWiweka*, Aniati Murni**; * PUSBANGJA LAPAN Email : wiweka@yahoo.com ** Fakultas Ilmu Komputer Universitas Indonesia, E-mail: aniati@cs.ui.ac.id
