Deep Learning-Powered Automated Road Damage Detection Using UAV Images

Puchha Aishwarya, Sappa. Geethika, RajapuLeela Vara Prasad, Pala Chandra Shekhar Naidu, Mrs. B. Kusuma Kumari

doi:10.62643/

Authors

Puchha Aishwarya, Sappa. Geethika, RajapuLeela Vara Prasad, Pala Chandra Shekhar Naidu, Mrs. B. Kusuma Kumari Author

DOI:

https://doi.org/10.62643/

Keywords:

Road Damage Detection, UAV, Deep Learning, CNN, Transfer Learning, Smart Infrastructure, Image Classification

Abstract

Road infrastructure maintenance is challenged by frequent damage from environmental factors and heavy traffic. Traditional manual inspection is time-consuming, labor-intensive, and error-prone. This paper presents an automated road damage detection system using Unmanned Aerial Vehicle (UAV) images and deep learning. High-resolution UAV images are processed to detect cracks, potholes, and surface deformations using a CNN-based model with transfer learning. Data augmentation techniques enhance performance across diverse road conditions. The system achieves 94.7% detection accuracy on a dataset of 8,000 UAV images, outperforming traditional methods (SVM+HOG: 78.3%, basic CNN: 86.1%). The approach enables real-time, cost-effective, and scalable road monitoring, reducing inspection time by 85% compared to manual surveys. This research contributes to smart transportation infrastructure by providing efficient automated road damage assessment.