Accelerating Disaster Response with Deep Learning and Image Processing Techniques

Abstract

This study examines the use of artificial intelligence-based image segmentation and image processing techniques in disaster management. The aim of the study is to integrate artificial intelligence and image processing techniques for fast and effective response in disaster areas to facilitate disaster management. Our hypothesis is that fast and accurate analysis can be performed with high-performance and fast AI-based segmentation models and image processing techniques on images taken from UAVs and satellites in disaster areas. In this process, important regions in the images are identified by applying segmentation processes and masks are created. Then, using these masks, numerical results can be obtained with image processing techniques with low computational cost. Thus, it is aimed to make fast and accurate decisions in disaster management. In this study, popular segmentation models were compared and analyzed using 2343 images obtained from Floodnet dataset. The results show that the SegFormer model provides detailed damage analysis and contour area or connected component analysis, which can provide both detailed and numerically accurate results for disaster management. This study reveals that the use of image processing and artificial intelligence in disaster management can improve response processes by increasing operational efficiency. ©2024 IEEE.