Marine biofouling, known as the unwanted accumulation of living organisms on submerged surfaces, is one of the main factors affecting the operation, maintenance and data quality of water quality monitoring sensors. This can be a significant challenge for marine deployed infrastructure and sensors in water. When organisms attach to the mooring lines or other submerged surfaces of the sensor, they can interfere with the sensor's operation and accuracy. They can also add weight and drag to the mooring system, making it more difficult to maintain the desired position of the sensor. This increases the cost of ownership to the point where it becomes prohibitively expensive to maintain operational sensor networks and infrastructures. Furthermore, the analysis and quantification of biofouling is extremely complex as it is based on biochemical methods such as the analysis of pigments such as chlorophyll-a as a direct indicator of the biomass of photosynthetic organisms, dry weight, carbohydrate analysis and protein analysis among others. In this context, this study has developed a method to estimate biofouling quickly and accurately on different submerged materials used in the marine industry and specifically in sensor manufacturing like copper, titanium, fiberglass composite, different types of polyoxymethylene (POMC, POMH), polyethylene terephthalate glycol (PETG) and 316L-stainless steel. To do this, in situ images of fouling organisms were collected with a conventional camera and image processing algorithms and machine learning models trained were used to construct a biofouling growth model. The algorithms and models were implemented with Fiji-based Weka Segmentation software. A supervised clustering model was used to identify three types of fouling to quantify fouling on panels of different materials submerged in seawater over time. This method is easy, fast and cost-effective to classify biofouling in a more accessible and holistic way that could be useful for engineering applications.
Keywords: Biofouling; Image analysis; Image segmentation; Instrumentation; Machine learning; Materials; Supervised classification.
Copyright © 2023 The Authors. Published by Elsevier B.V. All rights reserved.