A Hybrid Deep Learning Framework for Automated Road Damage Detection Using UAV Images

Abstract

Road infrastructure plays a vital role in ensuring safe and efficient transportation, making timely detection of road damage essential for reducing accidents and maintenance costs. Traditional inspection methods are often slow, expensive, and dependent on manual observation, making them unsuitable for largescale monitoring. This project presents an automated road damage detection system that uses Unmanned Aerial Vehicle (UAV) images and deep learning techniques to identify defects such as cracks and potholes. High-resolution aerial images are captured using UAVs and processed through advanced YOLO-based object detection models to accurately locate and classify damaged road sections. The collected images are preprocessed to improve detection performance, and the trained model is evaluated using standard performance metrics. The proposed system provides faster inspection, minimizes human effort, and enables continuous monitoring of road conditions over large areas. The results demonstrate that UAV-assisted deep learning offers a reliable and efficient solution for modern road infrastructure maintenance and supports timely repair planning.