Intelligent Crack & Corrosion Detection in Infrastructure using Edge Computer Vision
Published:
Building on the autonomous inspection work, this project detects structural cracks directly from aerial imagery using deep learning deployed on edge hardware — enabling real-time assessment during flight rather than offline post-processing.
Output
- Reyes-Munoz, J. A., Ortega, A. G., Rizia, M. M. (equal contribution), Choudhuri, A., & Flores-Abad, A. Autonomous aerial system for intelligent close-quarter inspection, Int. J. Intelligent Robotics and Applications (2026, open access). Article
- Deep Learning based Aerial Intelligent Crack Detection in Infrastructure using Computer Vision at the Edge (preprint, 2022) — Authorea
The system achieved 98.44% mAP@50 for crack detection and 72.33% for corrosion, with onboard AI inference, and was validated in real industrial conditions including a 26 m power-plant stack.
Edge device used: NVIDIA Jetson Nano
