Sign In

Communications of the ACM

ACM TechNews

Show Me the Methane


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
Gas flare at an oil refinery.

University of California, Santa Barbara researchers developed a system that uses hyperspectral imaging and machine learning for more accurate detection of methane emissions.

Credit: Servais Mont/Getty Images

A new system developed by University of California, Santa Barbara (UC Santa Barbara) researchers uses hyperspectral imaging and machine learning for more accurate detection of methane emissions.

The researchers used data from hyperspectral cameras in the vicinity of the Four Corners region of the American Southwest, where methane emissions are high. Hyperspectral imaging can capture spectral “fingerprints” that correspond to methane's 2,200-2,400 nm spectral band.

Said UC Santa Barbara's Satish Kumar, "We used a deep learning model to train the computer to learn the shape that a methane gas leak takes as it is released and spreads," which helped to pinpoint the location from which methane was being emitted, and to differentiate between methane and other hydrocarbons in the same image.

From The Current (UC Santa Barbara)
View Full Article

 

Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA


 

No entries found