Drone inspection platform

Drone inspection platform

In early 2023, in collaboration with our partners at Power Drone, we developed an advanced solution to detect and prevent faults in Veolia’s heat transfer network in Warsaw, Poland Our innovative approach involved deploying drones equipped with infrared cameras.

Veolia manages an extensive heating network spanning thousands of kilometers, grappling with an aging and unreliable infrastructure. The outdated and imprecise pipe plans complicate the process of pinpointing and resolving frequent leaks, which not only disrupt services but also pose environmental risks.

Leveraging Meta’s Mapillary capabilities and other open-source tools, we developed a drone-centric inspection platform that seamlessly produces a street-view-like map of the heating system through the lens of the drone. Employing WebODM, we generated thermal orthophoto images, enabling inspectors to easily identify steam leaks. Additionally, we implemented a feature to trace the inspected route and annotate potential leakage points. Leveraging the power of machine learning, specifically sklearn, we trained models to autonomously annotate images with relevant information.

The result: reduced breakdowns, saved energy and water, and improved customer satisfaction for Veolia. This project showcases the transformative impact of drone technology on the efficiency and sustainability of district heating systems.

Drone inspection platform for Veolia’s heating system helping to prevent leakages and outage.

Business Challenge

Veolia operates a vast heating network of thousands of kilometers, with an old and unreliable infrastructure. The pipe plans are outdated and inaccurate, making it hard to locate and fix the frequent leakages that cause service disruptions and environmental hazards.

Solution

RSC leveraged Meta’s Mapillary and other open-source tools to create a drone-based inspection platform that provides a street-view perspective of the heating system. Using WebODM, RSC generated thermal orthophoto images that allow the inspectors to easily spot the steam leaks. Using sklearn, RSC trained machine learning models that automatically annotate the images with relevant information.

Expected Outcome

• Improve transparency and compliance with regulatory standards
• Cut down the cost and risk of manual and hazardous inspections
• Boost accuracy and reliability of leakage detection and prevention

Toolset

• Mapillary
• OpenDroneMap
• WebODM
• PostgREST
• Angular
• Scikit-Learn
• DJI Thermal SDK