Echo- and Image-Domain Fusion for Micro-Leak Detection and Localization in Subsea Gas Pipelines

The detection of microleaks in subsea gas pipelines remains challenging in shallow-water environments because weak bubble-plume echoes are often obscured by seabed reverberation, ambient noise, and platform-induced interference. This study develops a custom multibeam forward-looking sonar system and a dual-domain framework for acoustic detection and localization of underwater gas microleakage. Local statistical enhancement suppresses stable background interference and strengthens anomalous bubble echoes. Blind deconvolution, adaptive grid-based thresholding, and spatial clustering then improve target sharpness, extract candidate bubble-plume regions, and reduce localized false detections. This unsupervised framework requires no pre-collected training data. It was evaluated in 30 independent sea-trial groups conducted in Bohai Bay using air to simulate leakage. Six orifice diameters from 0.5 to 3.0 mm were tested under different compressor-indicated pressures. The plume-observation distances (defined as the sonar-to-leak-device distance at the first confirmed plume response in the sonar image) ranged from 32 to 66 m across the tested conditions. Under the minimum tested condition of a 0.5 mm orifice and a compressor-indicated pressure of 0.5 MPa, the plume was observed at a distance of 32 m. The results demonstrate the feasibility of the proposed system for ROV-assisted detection and localization of subsea gas microleakage in low-visibility shallow-water environments.