Generation of Detailed Classification Maps using High-Resolution Satellite Images At Country-wide Scale

A generation of detailed classification maps requires high-resolution satellite images classified into numerous classes depending upon the region of interest. This paper addresses the challenges encountered towards generating such classification maps at a large scale. While using a high-resolution satellite image at a large scale, one could face challenges during the pre-processing phase, starting with mosaicking satellite images, dealing with sample collection, projection issues, or even using classification algorithms to develop classification maps. Hence, this article provides details on such tasks performed at a larger scale. A CNN-based supervised method is used to train the model for the classification from the approximately equal size of samples of the interested land cover land use (LCLU) classes. In addition, few land-use classes such as parks and airports are used from Volunteered Geographic Information (VGI) sources.