A novel drone-enabled approach for gas leak detection using thermal image flow analysis

The recent development of gas imaging technologies has raised a growing interest in various applications. Gas imaging can significantly enhance functional safety by early detection of hazardous gas leaks. Moreover, optical gas imaging technologies can be used to identify possible gas leakages and investigate the amount of gas emission in industrial sites, which is essential, primarily based on current efforts to decrease greenhouse gas emissions worldwide. Therefore, exploring the solutions for automating the inspection process can persuade industries for more regular tests and monitoring. One of the main challenges in gas imaging is the proximity condition required for data to be more reliable for analysis. Therefore, unmanned aerial vehicles can be very advantageous as they can provide significant access due to their maneuver capabilities. Despite the advantages of drones, their movements and sudden motions during hovering can diminish data usability. This paper proposes a novel approach to localize the possible gas leak from an image stream collected by an aerial platform. The introduced solution localizes the areas that demonstrate a gas flow-like motion and the origin of gas emission using thermal image flow analysis. Moreover, we investigate the use of the phase correlation technique to reduce the effect of drone movement during hovering. The significance of the results presented in this paper demonstrates the possible use of this approach in the industry. (author)