Automatic geolocation and measuring of offshore energy infrastructure with multimodal satellite data

Ma, Ping and Macdonald, Malcolm and Rouse, Sally and Ren, Jinchang (2023) Automatic geolocation and measuring of offshore energy infrastructure with multimodal satellite data. IEEE Journal of Oceanic Engineering. ISSN 0364-9059 (In Press)

[thumbnail of Ma-etal-IEEE-JOE-2023-Automatic-geolocation-and-measuring-of-offshore-energy-infrastructure]
Text. Filename: Ma_etal_IEEE_JOE_2023_Automatic_geolocation_and_measuring_of_offshore_energy_infrastructure.pdf
Accepted Author Manuscript
License: Strathprints license 1.0

Download (1MB)| Preview


With increasing trend of energy transition to low carbon economies, the rate of offshore structure installation and removal will rapidly accelerate through offshore renewable energy development and oil and gas decommissioning. Knowledge of the location and size of offshore infrastructure is vital in management of marine ecosystems, and also for safe navigation at sea. The availability of multimodal data enables the systematic assessment of offshore infrastructure. In this paper, we propose an automatic solution for the geolocation and size evaluation of offshore infrastructure through a data fusion model of Sentinel-1 Synthetic Aperture Radar (SAR) data and Sentinel-2 Multi-Spectral Instrument (MSI) imagery. The use of the Sentinel-1 (SAR) data aims to quick localization of the candidate offshore energy infrastructure by its all-weather imaging capabilities, while the high-resolution optical data provided by the Sentinel-2 can enable more accurate localization and measurement of the offshore infrastructure. To be specific, a candidate detection model is applied to a time-series of Sentinel-1 images to extract the ‘guided area’ of the infrastructure, followed by morphological operation based precise localization within an individual Sentinel-2 image as well as estimating the size of each structure. With validation against the ground truth data of the Scottish waters from the baseline and closing bays, to the limit of the Exclusive Economic Zone of Scotland, an area of 371,915 km2, our method has automatically identified 332 objects with an omission error of 0.3% and a commission rate of 0%. Our proposed method was comprehensively compared to two state-of-the-art offshore energy infrastructure detection algorithms. The results validate that our method achieves the highest overall accuracy of 99.70%, surpassing the compared methods by 3.84-12.50%. For the size evaluation, the achieved mean topside area size error of oil/gas platforms and the mean error for diameter length measurement of wind turbines both are 1 pixel in Sentinel-2 images, providing an effective technique for the identification and estimation of offshore infrastructure.