Geological classification of hyperspectral imagery in Hongseong Area, South Korea

Abstract

Hyperspectral imagery can be applied to geological classification of the earth surface with few vegetation or bare soil. In Korea, about 65% of land surface is covered with forest and bare soil area on a large scale rarely exists, so it is difficult to classify surface geology by using traditional remote sensing techniques. In spite of these circumstances, bare land surfaces are distributed on a small scale and hyperspectral imagery of these places is obtained to perform sensitive geological classification. Hongseong area is located the west shore of Chungcheongnam-do, South Korea and there are several quarries. The area of quarries is enough to detect and to analyze their geological features by using hyperspectral imagery analysis. Hongseong area consists of Precambrian schist group, intrusive Cretaceous granite group and metamorphic rocks of unknown ages. The spectral angle mapper (SAM) and the mixture-tuned matched filtering (MTMF) methods were used to classify its geological features and the results from each method were compared each other. Both results demonstrate that hyperspectral imagery is useful to detect specific rocks for geological classification, and the MTMF method produced more accurate detection than the SAM method.