CLASSIFICATION OF SATELLITE IMAGES USING DEEP LEARNING FOR DISASTER MANAGEMENT

Abstract

Abstract— Satellite imagery plays a crucial role across diverse domains, encompassing agriculture, urban planning, disaster management, and environmental monitoring. The precise and effective categorization of satellite images is vital for extracting valuable insights and facilitating well-informed decision-making. In this investigation, we suggest leveraging artificial intelligence methodologies for the classification of satellite images. We compile a comprehensive dataset of labelled satellite images, representing various land cover types or objects of interest. Prior to analysis, the dataset undergoes preprocessing to enhance image quality, eliminate noise, and standardize the data. To augment the dataset and enhance the model's ability to generalize, we employ techniques such as rotation, scaling, and flipping. Future research avenues could involve exploring advanced deep learning architectures, including attention mechanisms or graph neural networks, to further elevate classification performance. Moreover, the incorporation of multi-sensor satellite data and temporal analysis holds promise for augmenting classification models, particularly in dynamic monitoring and change detection applications.