India is quickly digitizing and transforming into a more intelligent nation, including smart cities and automated power transmission and distribution networks rapidly integrating the grand scheme of things (smart metering, smart grids etc.). With about 6,000,000 villages across the country and 68% of the people living in rural areas, it is of the utmost importance to provide them with innovative technology to transform India into a smart nation. Several real-time applications, including monitoring environments, the military, surveillance, and geographic surveys, make substantial use of satellite image categorization. Thus, accurate satellite image categorization is required to improve classification accuracy. Classification of land cover relies heavily on the efficient implementation of satellite image classification using sentinel images. Recent techniques for deep learning facilitate the interpretation of temporal and spatial data for land cover categorization in remotely sensed locations. This study applies a classification approach utilizing a transfer learning model on Euro sat dataset to classify sentinel-2 images that yield in developing a thematic map. There are 27,000 categorized Sentinel-2 satellite images in the Euro sat collection, split over 13 spectral bands and ten classification levels. High computational and achievement are maximized and connected using data augmentation, early halting, and adaptive learning rates. Hence, the transfer learning model ResNet50 obtains 96% accuracy with a 90% to 10% ratio for splitting. Thus, the resulting categorization model is broadly accessible in observatory applications, which aids in developing the smart village.
Land Cover and Land Use Classification Approach to Maintain the Green Environment in Smart Village
09.08.2023
521592 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Radar Remote Sensing: Land Cover Classification
British Library Conference Proceedings | 1998
|A land cover classification for Vermont
NTRS | 1981
|NTRS | 1982