Construction of Jakarta Land Use/Land Cover dataset using classification method

Abstract

The field of remote sensing has drawn a lot of attention recently. However, collecting necessary ground truth data for research in this field requires a lot of effort. Therefore, this paper presents a method for constructing estimated ground truth data using classification. This method reduces the workload in collecting remote sensing ground truth data. The contribution of this paper is to prepare and provide estimated Land Cover/Land Use (LULC) ground truth data of Jakarta area using the proposed method. The estimated ground truth data then can be used along with remote sensing image of Jakarta area to form dataset, which can be used for remote sensing research. For the estimated ground truth data to be reliable, the employed classification model have to achieve a reasonably good result. This research compares several algorithms to find the classification model with the best result for this case. The experimental result shows that Neural Network with single hidden layer of 30 neurons achieves best test accuracy of 75.41%. The method of this paper has been successfully implemented to construct LULC dataset of Jakarta area.