Research Projects

India is rapidly urbanizing, with a significant portion of its population living in informal settlements. However, there is a lack of knowledge on the dimensions and patterns of these settlements, as well as the extent of greenspace coverage in cities. Satellite images can be used to monitor the urban environment periodically, but high-resolution images are not always available. Coarse/medium resolution images like Sentinel-1 and Sentinel-2 are available at excellent temporal resolution, allowing for frequent and affordable monitoring. Detecting urban parameters from these images is challenging, and existing methods have shown limited success. The proposed research activity aims to advance urban monitoring methods using freely available satellite images and deep learning techniques, particularly in the context of Indian urban areas. The project will extract several important urban indicators from these images, and develop unsupervised and domain adaptation-based methods to reduce the requirement of annotated data. The project will generate urban/non-urban maps and building density maps for India at regular intervals, as well as urban change maps. Unsupervised semantic segmentation approaches will be used to generate population estimation for India based on satellite images and social media data. Additionally, maps indicating the prevalence of urban green vegetation will be generated to indicate a city's health. Upon successful completion, the project will create a low-budget, spatially accurate, and periodically updated urban monitoring system for India.