Feeding in sea-based aquaculture is typically either on a strict schedule or controlled by an operator monitoring the population through the use of cameras. The operator is thereby in charge of reducing food waste and optimising growth by feeding when the fish are hungry and stopping when sated. The use of experienced personnel and cameras has, to some degree, reduced both overfeeding and underfeeding. However, cameras have a limited observational volume, and this thereby limits the information source for the personnel. The reliance on subjective decision-making also increases the likelihood of human error considerably. Due to this, over- and underfeeding still occur.
The OptiFe3d project will develop a new system based on multi-beam sonar technology which has not been previously utilised in the Norwegian Aquaculture industry for feed optimisation. The multi-beam sonar will be used to perform behavioural analysis and pellet detection, which, together with environmental data, will provide a basis for feed recommendations. Compared to cameras and single-beam sonars, the multi-beam sonar has a larger observational volume, can separate very small objects from each other, manage extremely accurate biomass estimation and build a live 3D “digital twin” of a net-pen population. Utilising this advantage, it will be possible to analyse salmon feeding behaviour on a group and individual level. The increased resolution also opens the possibility for accurate pellet detection.
By combining feed behaviour and pellet loss detection, the project will optimise feeding to avoid both over and underfeeding, saving both financial resources and reducing the industries’ carbon footprint, ultimately benefiting the entire aquaculture industry.
Feeding is one of the core operations of fish production, however nearly 10% of the yearly feed consumption in Norwegian Atlantic salmon production is lost to the environment. This is a major environmental and financial concern for the industry. OptiFe3D’s main goal is to provide an efficient and accurate decision support system (DSS) to optimize the feeding operation based on new methods that correlate environmental conditions and salmon’s appetitive behaviour alongside pellet loss estimation. These new methods adopt state-of-the-art Machine Learning (ML) and algorithms supported by ground-breaking 3D-sonar technology.
OptiFe3D will focus on the following research areas (RA): RA1: 3D monitoring and detection of fish feeding behaviour, RA2: Autonomous pellet loss estimation, RA3: Using behaviour and pellet loss information to optimize feeding.
Environmental and biological monitoring will be performed at two separate Lingalaks sites to ensure versatility of OptiFe3D's DSS. Environmental monitoring is carried out using Aanderraa's instrumentation, and long-standing expertise in the field, to identify environmental drivers that might influence the salmon's behaviour and pellet distribution. Aquabio's unique high-resolution 3D-sonar will be utilised for the biological monitoring of both fish behaviour and pellets.
The biological and environmental data forms the basis for the development of the ML algorithms for detection of fish appetitive behaviour, at both group and individual level (RA1), and pellet distribution (RA2).
Lastly these algorithms will be integrated to create a proof of concept for the OptiFeed's DSS (RA3).
OptiFeed with this novel decision support system will provide recommendations for duration and intensity of feeding thereby aiding personnel in eliminating food waste. The intelligent feeding system will also assess the quality of each feeding period, while challenging the current feeding strategies with brand-new technologies and methods.