Automated Leak Detection Using Machine Learning and Multi-Platform Remote Sensing
(Room Conference Hall)
04 Apr 18
2:00 PM
-
3:30 PM
This paper discusses the development of the Smart Leak Detection (SLED) system, a multi-platform remote sensing technology for the autonomous detection of liquid hydrocarbons, and recent work funded through the U.S. DOE to extend this capability to the detection of fugitive methane emissions. The technology is suitable for both mobile platforms (manned and unmanned aircraft, land vehicles, etc.) and stationary platforms, such as fixed installations at pump stations and block valve sites. SLED fuses inputs from various types of sensors and applies machine learning techniques to reliably detect “fingerprints” of liquid and gas leaks in near real-time. Leak characterization was performed by imaging different types of hazardous liquid and gas (crude oil, refined products, methane) in several different environmental conditions and sceneries. In addition to detecting land (pipeline) leaks, SLED has also been extended to detect marine oil spills using fused visible and SAR satellite imagery.