Development of artificial intelligence-based methods for disaster management and situation analysis from remotely obtained

Various natural disasters have occurred in every period since the world was founded, and these disasters have caused serious loss of life and property. With the development of deep learning (DL) networks, the impact of any disaster can be analyzed and the affected areas can be identified. Regardless of the type of disaster, the damage it causes can be serious. Immediately identifying the areas where damage has occurred is of great importance in preventing loss of life and property. For this purpose, unmanned aerial vehicles (UAVs) have recently begun to be used in search and rescue operations. The flexibility of UAVs, their ability to move and quickly position themselves, and their low cost have made them preferred in search and rescue activities. For more successful results, incorporating Deep Learning (DL) models into unmanned aerial vehicles may be a good method. Two separate data sets were used in this study. A new transfer learning method based on convolutional block attention mechanism is proposed on our first dataset, UAV-based air disaster image dataset (AIDER).The proposed method focuses on the rapiddetection of natural disasters. At this stage, the aim is to provide accurate and timely information for urban planning and disaster management. A set of methods are proposed using Deep Neural Networks on our second dataset, High Resolution Aerial Imagery Dataset for Understanding Post-Flood Scene (FloodNet). Transfer learning based image classification was used to distinguish flooded areas from undamaged areas. Semantic segmentation methods were used to detect flooded areas, buildings and roads, and to distinguish between natural water and flood water. In the final stage, scenes were made meaningful with visual questions and answers. The effectiveness of deep learningbased automation systems in accelerating post-disaster response processes and optimizing human intervention is emphasized.