摘 要 ：The earth’s surface landscape classification was obtained by computer analysis of LANDSAT multispectral data. Clustering analysis of spectral data provides an element, repeatable classification based upon spectral characteristics of earth’s surface landscape. The classification procedure repeated until the data are classified into the natural lusters. In this experiment, the clustering algorithm yields 17(0–16) types of earth′s surface landscape by 8 iterations.The accuracy of the classification was verified by a comparison with aerial photography and ground investigation, and the agreement was in the order of 83%. The results from automatic classification were satisfactory.
Abstract:The earth’s surface landscape classification was obtained by computer analysis of LANDSAT multispectral data. Clustering analysis of spectral data provides an element, repeatable classification based upon spectral characteristics of earth’s surface landscape. The classification procedure repeated until the data are classified into the natural clusters. In this experiment, the clustering algorithm yields 17(0–16) types of earth′s surface landscape by 8 iterations.The accuracy of the classification was verified by a comparison with aerial photography andground investigation, and the agreement was in the order of 83%. The results from automatic classification were satisfactory.