DESIGN AND IMPLEMENTATION OF DISASTER IMAGES FROM SATELLITE IMAGES USING VGG-NET TECHNIQUE

Abstract

Abstract - In recent times, the application of deep learning techniques in disaster monitoring has shown remarkable potential, particularly with the abundance of satellite imagery available. This study proposes a comprehensive framework for disaster monitoring employing satellite image processing and deep learning methodologies. The proposed framework integrates various stages including input image acquisition, pre-processing through median filtering, segmentation via Fuzzy C-Means (FCM) algorithm, feature extraction utilizing Gray-Level Co-occurrence Matrix (GLCM), Convolutional Neural Network (CNN) modelling, and finally, output generation. Initially, raw satellite images are acquired and subjected to median filtering to enhance their quality and reduce noise interference. Subsequently, FCM segmentation is applied to partition the images into meaningful regions, facilitating better feature extraction. GLCM is then employed to extract texture features from the segmented regions, enabling the characterization of various disaster-related patterns. The extracted features are fed into CNN (VGG-NET) architecture for learning and classification. The performance of the proposed framework is evaluated using several performance metrics, including accuracy, specificity, sensitivity, and precision. The experimental results demonstrate promising performance, with an accuracy of 95.215257%. Additionally, the framework achieves high specificity (96.899446), sensitivity (96.057352), and precision (96.478399), indicating its effectiveness in accurately identifying disaster-related features in satellite imagery. Overall, the proposed deep learning-based approach presents a robust solution for disaster monitoring, leveraging the power of satellite image processing and convolutional neural networks. The achieved performance metrics underscore its potential for real-world applications in disaster management and response systems. Keywords: Deep learning, Disaster monitoring, Satellite imagery, Image processing, Median filter, Segmentation, Fuzzy C-Means (FCM), GLCM (Gray-Level Co-occurrence Matrix), Convolutional Neural Network (CNN), etc.