Separability Analysis and Classification of Rice Fields using KOMPSAT-2 High Resolution Satellite Imagery

Global warming has become one of the world's most serious social issues. Nitrogenous fertilizer, which is generally used for agricultural activity, is a major source of greenhouse gas in the atmosphere. Investigating the distribution of crop fields is thus essential to controlling greenhouse gases. Analyses using remotely sensed data such as aerial photos and satellite images can be more cost-effective and reliable for large agricultural areas. Classification techniques that use multi-date satellite images are efficient tools for identifying agricultural land. In this study, image co-registration between images from different dates is performed for preprocessing. The Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) atmospheric correction technique is applied to convert radiance data in the satellite image to reflectance data. We then analyze and compare several vegetation indices using separability analysis. Rice fields are classified using the composition of the multi-temporal satellite image and the most efficient vegetation index. The extracted rice fields are compared with reference data derived from aerial photo interpretation. The classification results provided agricultural accuracies of approximately 90%. In addition, the extracted rice fields had acceptable producer and user accuracies of 83% and 78% respectively in Daejeon and 80% and 72% respectively in Gongju.