Drones to Cut Road Inspections
Industry
Transport
Client
Private client
Team
SCALE
Year
2024
We were hired to deploy AI algorithms for drone squads to monitor roads continuously and eliminate months-long inspection delays.
Challenge
Inspecting city roads and transport infrastructure took 6–8 weeks per cycle with manual crews, causing delayed repairs and safety risks. The client needed a faster, automated way to detect road damage and incidents in real time.
Approach
Deployed autonomous drones with AI vision (YOLOv10) for 24/7 road monitoring, streaming data via a 5G base station network.
Implemented a real-time image recognition pipeline (thermal + RGB) to detect potholes, cracks, and traffic violations, using multi-object tracking (MOTA, IDF1) metrics to ensure accuracy.
Built an automated battery swapping mechanism for drones, enabling continuous operation beyond normal battery limits.
Integrated a geospatial mapping module to pinpoint issues for city repair crews, and set up alert dashboards for transport authorities.
Architecture/Backend
RGB + video via a 5G + real-time image-recognition pipeline; geospatial mapping.
AI/ML models
YOLOv10-based object detection for persons/vehicles on RGB.
Infrastructure/Deployment
5G low-latency comms between UAVs and ground stations; continuous autonomous operations.
Key results
Inspection cycle time reduced from ~6–8 weeks to just hours for coverage of pilot routes. This enabled same-day identification of road defects and hazards (vs. nearly two-month lag previously).
~65% reduction in operational inspection costs (estimated) in the pilot, by cutting down on manual patrols and associated labor/vehicle expenses (confirmation pending).
Improved public safety and responsiveness: issues were flagged instantly, accelerating repair scheduling (city crews now respond within a day vs. weeks).
