On the Autonomous Inspection and Classification of Marine Growth on Subsea Structures

Marine growth challenges the structural integrity of offshore facilities due to increased hydro dynamical loads. As a consequence, marine growth cleaning on offshore structures has been performed for many years. While the industry has shifted from diver-assisted to cleaning driven by remotely operated vehicles, the process remains costly and ineffective. This paper explores the possibilities for introducing an increased level of automation for marine growth inspection and classification. Specific attention is given to sensor technologies and methods for constructing a 3D representation of the offshore structures in order to assess the thickness and composition of marine growth. While optical-based methods show positive potential further work is needed to investigate the robustness to flicking sunlight and turbidity issues experienced in areas close to the water surface. The review of classification methods reveals several promising approaches where deep learning is applied for the categorization of marine growth. The training relies on large databases of relevant images which are not currently available for marine growth on offshore structures. Further work is needed for investigating if virtual images can be used in combination with a reduced set of real images.