Automated Rapid Post-Disaster Assessment Using Uncrewed Aerial Vehicles

Nima Shirzad-Ghaleroudkhani, Danial Jozi, Melissa Milena Trigueros Bedia, Shaghayegh Abtahi, Maziar Jamshidi, Mustafa Gül

Keywords: Post-disaster Assessment, UAV Imagery, Computer Vision, Image Processing, Naïve Bayesian Classifier

TL;DR: An AI-assisted framework is presented for rapidly detecting and classifying residential building damage using only post-disaster images captured on UAVs.

Abstract: Climate change is intensifying the frequency and severity of natural disasters, such as hurricanes and earthquakes, resulting in devastating human and economic losses and creating a growing demand for rapid, accurate post-disaster assessment. Traditional ground-based inspections are often time-consuming, resource-intensive, and delayed due to accessibility challenges, while many existing image-based methods depend on pre- and post-disaster data that are not always available. To address this gap, this study proposes a novel framework for rapid and automated residential damage assessment using only single post-disaster UAV images. The framework operates in two stages: first, residential houses are automatically detected and identified and mapped from UAV-recorded video footage using a YOLOv8 segmentation model combined with DeepSort object tracking. In the second stage, a feature-based classification approach enhances damage detection by extracting texture- and edge-based features from house images, constructing four indices that are evaluated using a Naïve Bayesian Classifier to estimate damage probability. Validation with real-life post-disaster UAV imagery demonstrates the effectiveness of the proposed method in distinguishing damaged from undamaged houses, underscoring the potential of UAV-based imagery and advanced AI techniques for rapid, scalable, and accurate post-disaster assessment.

Submission Number: 20