SIMULATION OF HYPERSPECTRAL IMAGE USING MULTISPECTRAL IMAGE FOR LAND COVER MAPPING

摘要

Land cover map is useful reference data on various research area such as environment, water resource, climate monitoring. Almost of land cover map was produced using multispectral imagery which has a few classes and low accuracy due to limitation of number of band. Hyperspectral imagery could be a powerful data to classify land cover type which can increase number of class and classification accuracy with more than 100 bands. However, unfortunately hyperspectral imagery cannot have global or regional coverage due to narrower swath width than multispectral imagery. In this study, we try to simulate hyperspectral image using multispectral image and spectral library from hyperspectral image. Simulated hyperspectral spectra shows similar pattern and absorption features with original hyperspectral data. Classification accuracy of simulated hyperspectral image also higher than multispectral image for 14 classes land cover mapping. Therefore, simulating of hyperspectral image using multispectral image could increase classification accuracy for detail land cover types in regional scale.