(A) study on the construction of an image processing system for remote sensing

Abstract

In this thesis, an attempt is made to develop land use classification based on clustering approach, which is patterned after the ISODATA technique, by remote sensing using Landsat and operations for land use classification are implemented sequentially. These operations include contrast stretch enhancement, image registration, principal component analysis, clustering, and classification. In clustering, all required parameters are set internally from the actual Landsat data, thus eliminating a priori estimates, initial clusters are chosen reasonably, and resultant clusters are determined easily by relying more heavily on the lumping of clusters. Detailed descriptions of the operational procedure are given and result of application of the clustering algorithm to Seoul area Landsat image is included.