Cultivated seaweed is the fastest-growing aquaculture sector worldwide and a multibillion-dollar industry. Monitoring environmental conditions (e.g. temperature, salinity, nutrients and irradiance) and biofouling organisms in a seaweed farm is important for making decisions related to growth optimisation. Deployment of underwater robots for line inspection, in combination with “deep-learning” approaches, have the power to provide a fast and reliable estimation of seaweed biomass. Here, we show state-of-the-art, yet cost-effective, and scalable technologies aimed at optimising monitoring in a Norwegian kelp farm. Robotic monitoring of kelp farms, including biomass growth, was assessed using a mini, cost-effective, remotely operated vehicle (ROV). For a fast and reliable estimation of kelp biomass, a robust set of images to build a data-centric machine learning platform was collected, where we developed computer vision applications supported by AI algorithms. Preliminary results from the MoniTARE project showed that there is a strong correlation (R²=0.85) between the ground-truth biomass (manually collected) and the biomass inferred through 2D computer vision techniques from recorded images. We also propose low-cost technological ideas for frequent monitoring of larval concentrations of bryozoans in the water column to assist in early fouling detection. Automation of kelp farm monitoring has the potential to revolutionize the industry by offering scalability of production and improved yield predictions.

Primary Presenter: Glaucia Fragoso, NTNU (glaucia.m.fragoso@ntnu.no)

Authors:

Martin Overrein, NTNU  (martimov@stud.ntnu.no)

Phil Tinn, MIT  (ptinn@mit.edu)