Research Article
Scalable Deep Learning for Categorization of Satellite Images
@INPROCEEDINGS{10.1007/978-3-031-81168-5_1, author={C. Lokanath Reddy and T. Mukthananda Reddy and M. Mahendra Reddy and M. Mohan}, title={Scalable Deep Learning for Categorization of Satellite Images}, proceedings={Broadband Communications, Networks, and Systems. 14th EAI International Conference, BROADNETS 2024, Hyderabad, India, February 16--17, 2024, Proceedings, Part I}, proceedings_a={BROADNETS}, year={2025}, month={2}, keywords={Hyperspectral remote sensing deep neural networks satellite images Python Google Collab are some of the index phrases}, doi={10.1007/978-3-031-81168-5_1} }- C. Lokanath Reddy
T. Mukthananda Reddy
M. Mahendra Reddy
M. Mohan
Year: 2025
Scalable Deep Learning for Categorization of Satellite Images
BROADNETS
Springer
DOI: 10.1007/978-3-031-81168-5_1
Abstract
The analysis of satellite data has become increasingly challenging due to the vast abundance of satellite photos available in recent years. Understanding and extracting valuable insights from these images necessitates a comprehensive grasp of the underlying data they portray. The capability to identify and categorize objects within satellite photos holds significant importance across various domains, including land planning, ecology, military intelligence, and ocean monitoring. With their wealth of spatiotemporal information, satellite images serve as invaluable resources for global remote sensing applications aimed at addressing a wide spectrum of issues. This study aims to investigate the complexities associated with analyzing satellite data by developing a specialized workflow focused on mapping streets and highways to monitor urban development in cities. The study emphasizes addressing learning challenges through the configuration execution, and evaluation of deep neural network experiments. To achieve this objective, publicly accessible methods and information are utilized. The data acquisition pipeline incorporates preprocessing techniques to effectively handle inputs with varying sizes, resolutions, and spectral channels. Despite the significant potential of satellite imagery, its widespread dissemination is hindered by various challenges, including issues related to data distribution, volume, quality, and accessibility. These obstacles further complicate the study of satellite images. Additionally, satellite imagery finds application in monitoring oceanic and geographical data, highlighting its diverse utility. The proposed strategy is anchored in a scalable end-to-end approach for interpreting satellite imagery, aiming to overcome the challenges associated with analyzing large-scale satellite datasets efficiently. Through this study, we aim to contribute to ongoing efforts in harnessing the power of satellite imagery for addressing global challenges and fostering sustainable development.
