Introduction
Global fish consumption has increased at an average annual rate of 3.1% from 1961 to 2017, which ranks higher than all other animal protein foods (meat, dairy, milk, etc.) That puts tremendous pressure on marine ecosystems, as it equates to roughly 180 million tons of fish species being consumed each year [1]. The aquaculture and fish-farming industry has been gaining traction as a commercially viable, sustainable and eco-friendly path forward.
In aquaculture explorations, fish and other aquatic species are bred, nurtured, harvested, and pro- cessed under controlled conditions designed to maximize quality and yield and minimize costs and environmental impacts. Traditionally, indoor aquaculture tanks rely heavily on periodic human inspection and operation. Technological innovation brought forward by aquaculture engineer- ing comes in response to the demand for even tighter control of water quality parameters, more efficient monitoring of fish larvae growth, and more automated “intelligent” feeding processes.
The most studied tasks requiring automation typically revolve around detecting fish size charac- teristics [4][5] expanded on this notion, developing a robotic system equipped with a camera and two laser projectors capable of directly estimating in real-time the biomass present in a large fish tank, with between 10% and 17% of relative error in biomass volume in a real aquaculture environment.
When it comes to fish biomass estimation in specific, estimates are crucial for evaluating fish growth during its growth cycle in the hatchery/nursery stages. The statistical characterization of a given’s tank population is fundamental for aquaculture explorations to control and adjust food and medicine dosage, to be able to detect fish population loss in early stages, and perhaps most importantly, to adequately monitor growth rates to decide the best timeline for financial decisions. Further advancement of economies of scale in industrial aquaculture nursery/hatchery scenarios are thus heavily dependent on the automation of monitoring and quality assurance processes.
To address this problem, this paper will present the Feedfirst Intelligent Monitoring System for Indoor Aquaculture Tanks, a system developed hand-in-hand with industry partners designed with three goals in mind: (i) continuous monitoring of water quality; (ii) continuous monitoring of larvae growth and population size; and (iii) automation and optimization of daily feeding.
Feedfirst Intelligent Monitoring System
The Feedfirst system was tailored to industry partners requirements in indoor aquaculture scenar- ios and the ultimate design goal is to close the perception-actuation loop, i.e. automate feeding control in accordance to perceived fish growth parameters. All measures from water quality to image data are relayed to a central computer that is responsible for both processing the incoming data and providing a user-interface to the tank operator (see fig.3).
The machine vision processing pipeline is intended to provide two important metrics: the count of population size and the size of each larvae fish. To achieve this, the submerged cylinder possesses 2 stereo-mounted cameras synced with an illumination system. Red light wavelengths are used so as to keep environmental disturbance to a minimum. Acquired images are subject to an image processing pipeline as shown in fig. 2. The prototype system was tested in operational conditions in a indoor aquaculture in Portugal.
References
[1] The State of World Fisheries and Aquaculture 2020. URL: ca9229en/CA9229EN.pdf.
[2] Mingming Hao, Helong Yu, and Daoliang Li. “The measurement of fish size by machine vision-a review”. In: International Conference on Computer and Computing Technologies in Agriculture. Springer. 2015.
[3] R Froese. “Short Communication Length-weight relationships for 18 less-studied fish species”. In: J. Appl. Ichlhyol 14 (1998).
[4] Flávio Lopes et al. “Structured light system calibration for perception in underwater tanks”. In: Iberian Conference on Pattern Recognition and Image Analysis. Springer. 2015.
[5] F Lopes et al. “Fish farming autonomous calibration system”. In: OCEANS 2017-Aberdeen. IEEE. 2017.