Development of a plume identification algorithm for optical gas imaging of natural gas emissions that requires no human intervention
Date
2020
Authors
Martinez, Marcus M., author
Zimmerle, Daniel, advisor
Marchese, Anthony, advisor
von Fischer, Joe, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
Recent growth in natural gas production in the United States has increased focus on reducing greenhouse gas emissions from the natural gas supply chain. Methane, the primary constituent of natural gas, is also a potent greenhouse gas. Optical gas imaging (OGI) is frequently used for emission detection in upstream and midstream sectors of the natural gas supply chain. Current OGI methods typically use mid-range infrared video cameras tuned to absorption lines of light hydrocarbons to make natural gas emissions visible to human operators. Prior studies of camera output have used human interpretation to determine if an emission is visible in the video stream, making it difficult to standardize measures of visibility between tests or to automate large test suites. This work presents a signal processing method which separates the background scene from the gas plume when used in controlled test conditions where video is collected in both leaking and non-leaking conditions. The method utilizes a novel frequency-based method that detects the high-frequency motion of the gas plume in the video stream. After background removal, the size of the gas plume can be quantified by thresholding the detected plume and measuring its size relative to the camera's field of view. The resulting metric eliminates the need for human evaluation of video streams. To demonstrate application of the method, multiple cameras were used to develop a relationship between emission rate and plume visibility over a range of viewing distances. Tests were conducted at the Methane Emissions Technology Evaluation Center, on CSU's Foothills Campus, using six identical OGI cameras (FLIR G300a camera cores with 38 mm lenses) to image the emission from multiple directions at a range 1 to 6 m. Gas was released from a mock well head at 17 to 196 g/h, with wind speeds of 1.8 to 3.0 m/s. Comparison with expert evaluation was used to set and validate the threshold levels; a 90% probability of detection requires a plume covering at least 13.8% of the camera's field of view. Testing indicated a linear relationship between emission rate and plume coverage fractions at a distance of 1 to 2 m, regardless of the viewing angle. Beyond 2 m, plume coverage drops rapidly, approaching the noise floor. While test conditions were limited, sufficient data was collected to demonstrate method functionality and its applicability to evaluating OGI emission detection systems.
Description
Zip file contains data spreadsheet and supplementary videos.
Rights Access
Subject
imaging distance
viewing angle
OGI
frequency-based