Fugitive emissions detection from infrared drone footage
Computer vision shown able to take over the manual interpretation of infrared footage used to find fugitive emissions.
Challenge
A large multinational engineering firm used infrared cameras to visualise fugitive emissions at industrial sites, with subject-matter experts interpreting the images by hand. It asked whether machine learning could replace that interpretation.
Solution
Using computer vision and deep learning on raw drone footage, Arundo demonstrated early results showing the interpretation can be automated given sufficient high-quality data, with the follow-on analysis and processing able to run on the Arundo platform.
Impact
Feasibility shown; the manual interpretation step can be automated.