Abstract
In supervised classification of land cover, ground truth data are used as training data for classification, or for post-classification validation of data. In classifying a large area, it is difficult to collect enough ground truth data. This study proposed a method of collecting ground truth data using both existing land cover data and existing land use data. By combining these two types of data, more reliable regions can be extracted as ground truth data. Using ground truth data collected by this method in China, a larger region was classified as a trial. The classification accuracy was 81% when the results were evaluated using the ground truth data. For the practical use, combining the ground truth data collected by the proposed method with those collected by the other method is recommended to enrich ground truth data inventories.
