Localised Land-Use Classification Using U-Net and Satellite Imaging

The existing cities are growing and it is becoming increasingly tough to maintain them. As more and more populations migrate to cities, growth is inevitable, which leads to improper development of cities. With increasing urbanisation and the advancement of technology, we get an opportunity to explore the idea of land-use classification. Modern software and models do provide such information but at a high zoom level which defeats the purpose of localised growth of an area. Using CNN techniques such as U-Net, we provide a comprehensive study on the results of U-Net and QGIS for land-use classification for higher zoom levels. This aids in directing the city's growth in the right direction, evenly distributing the expanding population with all of the necessities, and concurrently maintaining the natural resources. We provide a U-Net model that performs exceptionally well in Indian conditions. The solution provided works great for zoom level for cities as well as localities in a city, which can be very useful for proper planning. We have also integrated the U-Net model with QGIS to detect roads and obtain data on the various types of buildings that are present in the area, which will assist our model to identify and forecast potential growth hotspots.