Study Of A Semantic Segmentation Algorithm For Disaster Assessment In An Edge Computing Environment

Abstract

When a natural disaster strikes, it often causes widespread destruction, making timely and accurate information essential for effective response and recovery. Traditional assessment methods, which rely on manual interpretation of aerial and satellite imagery, are limited by delays, high operational costs, and a high risk of human error. This project aims to enhance disaster assessment by leveraging deep learning and edge computing to enable fast, reliable analysis during emergencies. The primary objective is to develop a system capable of accurately identifying objects and classifying them based on the amount of reported damage. This system will then be adapted for real-time, automated semantic analysis on edge devices such as drones and IoT sensors. The goal is to minimize reliance on remote data centers, improve on-site decision-making, and support efficient disaster response in resource-constrained environments by balancing accuracy with computational efficiency.