Fishful thinking

Share this on social media:

Tim Reynolds on some of the ways imaging is used to farm fish

Seaweed is grown on ropes, the moorings and position of which can be tracked by a vision system. Credit: PEBL

​​Around half of all fish consumed globally are now farmed. The amount of wild fish caught has remained at roughly 90 million tonnes per year since the 1990s, with the United Nations Food and Agriculture Organisation estimating that one-third of global wild stocks are already being fished beyond sustainable limits. In the case of salmon, almost three-quarters are farmed, with major production centres in Norway, Scotland, Chile and Canada. Farms are looking for ways to improve productivity, but there are significant concerns over the environmental impact caused by waste feed and the transmission of diseases to native fish populations. Imaging can help with both, making farms more productive and reducing their environmental impact.

‘The industry’s biggest expense is feed,’ said Leif Johannesen, head of R&D at JT Electric. ‘Optimising feed can minimise costs and have a big impact on environmental issues, reducing the quantity of wasted feed.’

JT Electric is based in the Faroe Islands and supplies electrical products to the aquaculture sector – a term that covers fish farming but also the farming of other aquatic organisms like shellfish and algae. JT Electric’s FishFeeder system uses cameras – including an upward facing camera – deployed in and around fish pens that may be 50 metres deep.

‘The system uses computer vision algorithms to analyse images coming from cameras in the pens,’ explained Johannesen. ‘The software can differentiate between pellets and fish, analyse patterns on screen and feed back to the system.’

Using the system, operators can adjust the feed mechanism to minimise ‘drop through’ of feed, maximising feed conversion rates.

Johannesen said the computer vision system is there to help get the best out of the facilities. Even a small increase in feed efficiency can make a big difference to a fish farm’s bottom line and environmental impact.

Digital aquaculture

As for most industrial sectors, digitisation is a hot topic along the supply chain from fish farm to fork. The aquaculture sector is in the early stage of transformation when it comes to technologies such as artificial intelligence, big data, machine learning and vision analysis. The industry needs to develop an understanding of where, how and if new digital technologies can be used.

Johannesen pointed out that many farms are in remote areas and subject to bad weather, requiring robust digital communications infrastructure. Full digital developments may be more rapid in land-based fish farms where communications are guaranteed.

Nevertheless, some remarkable developments have been achieved. Aquaculture start-up Aquabyte is selling machine-learning technology that claims to improve fish welfare and overall farm efficiency.

Started in his bathtub by CEO Bryton Shang, Aquabyte is now an established software company with a global presence.

Writing on the company blog, Shang described how he adapted the machine-learning technologies he had previously applied to stock market trading and cancer diagnostics.

In the bathroom of his home in San Francisco, he was able to measure the distance to robot fish and develop a 3D model that successfully estimated their weight using two cameras. But a Californian bathtub is a world away from the real-life conditions of a fish farm pen with tens of thousands of fish.

Combining the innovation culture of Silicon Valley with expertise from the Norwegian aquaculture sector, Aquabyte has developed an automatic weight measurement system, enabling daily, continuous measurement of several thousand fish with an accuracy of 98 per cent.

This close monitoring of fish growth and health enables farms to provide more precise slaughter estimates.

The company has also developed automated lice counting software that runs with the same imaging hardware. Sea lice are not harmful to humans, but the lesions they cause, even in minor infestations, can render the salmon unmarketable. Severe infestations can lead to secondary infections and even mass mortality in the farm.

The global economic impact of sea lice infestations on the industry is of the order of $500 million annually. And because sea lice also affect local wild fish stocks, the issue has a knock-on environmental impact.

Aquabyte’s lice counting application enables producers to get a better, more accurate overview of the sea lice situation in their pens without increased handling of the fish.

Injection at pace

Another aspect of fish health and welfare in aquaculture is immunising stock against disease. Vaccinating large numbers of fish – usually by immersion or injection – is a challenging task.

The injection route is a highly skilled manual process, but new mechanised systems can deliver a combination of vaccine formulations accurately and consistently. These are able to easily export all data on the vaccinations delivered, size and distribution.

For example, the Pharmaq Fishteq range of equipment uses a machine vision system to determine the inoculation site on each target fish. The image recognition software can cope with a wide range of fish sizes (120 to 250mm or 20 to 150g) and a variety of species. It can determine precisely where the injection site should be to within +/- 0.3mm, with a dose volume accuracy of +/- 2 per cent. The machine can deliver up to three vaccines simultaneously at two different injection sites.

All the Pharmaq models can vaccinate up to 8,500 fish per hour. This minimises extra handling of the fish, leading to improved fish welfare and production economics.

Filleting inspection

Vision systems have multiple applications along the fish processing chain. The Maritech Eye, from Norwegian software firm, Maritech, is designed for automated quality control of fish fillets. It was launched in 2020 and is the result of a collaboration with HySpex, the hyperspectral imaging arm of Norsk Elektro Optikk (NEO) and Nofima, the Norwegian food research institute.

‘The relationship with NEO started in 2003 to 2004,’ recalled senior Nofima scientist Karsten Heia. ‘NEO were producing high-end hyperspectral cameras – mainly for military applications – but they were affordable and therefore applicable for industrial uses.’

The Maritech Eye system uses hyperspectral imaging to gauge the quality of filleted fish. Credit: Maritech and HySpex

Heia continued: ‘The system is based on Nofima research funded by the Norwegian Seafood Research Fund to increase the export value of the fishing sector, finding new solutions for aquaculture and wild fish.’

The original application analysed blood content in unprocessed whole white fish, determining quality before cutting. To do this it used a Baldur hyperspectral camera from HySpex to measure the amount of blood in white fish fillets – a marker for quality grading.

The device can be adapted for other applications. Just before Christmas last year, a Maritech Eye was installed at Mowi Consumer Products UK to enhance the quality assessment of its smoked salmon. It uses high spatial resolution to spot blood leading to dark marks – often rejected by smoked salmon consumers. The machine enables the grader to spot and sort at industrial speed, fully documenting the quality of each unique fish fillet processed. It moves quality assessment onto the production line, rather than relying on sampling and the uncertainty of offline assessment.

‘Recent applications we have worked on [using hyperspectral imaging] include identification of nematodes in white fish and assessing potential shelf-life,’ said Heia. The temperature profile history of a fish from net or pen to processing has a major impact on its shelf life. Monitoring the oxidation of haemoglobin in fish blood trains the system to determine the freshness of the fish – in particular, whether a breach of the cold chain has occurred.

Other uses might include analysis of fatty acid content in salmon, or using a diffuse reflectance light setup to assess for fish welfare indicators such as damaged or bleeding fins.

‘We are continually developing the camera to make it better, faster and more light sensitive,’ said Trond Løke, CEO of HySpex. ‘In the near future, we are extending our spectral range into the shortwave infrared up to 2.5µm wavelength.’

Løke added: ‘With our camera hardware and the [software] models from Nofima it is relatively simple to add new features through software updates, covering other quality parameters as required. It used to take years to develop, validate and verify a model - now it is much swifter.’

Per Alfred Nordaune Holte, vice president of technical solutions at Maritech, commented: ‘The development cycle for new applications can now be a matter of months. It is now a very structured approach.’

Not just fish

Of course, aquaculture is about more than fish. A Welsh company seeks to empower coastal communities using imaging to develop sustainable ways to cultivate, monitor and protect native seaweed species.

Plant Ecology Beyond Land (PEBL) is developing SeaLens, an autonomous monitoring system for small and medium-sized aquaculture organisations. Modular in design, it integrates sensors to monitor various parameters, such as seawater temperature, salinity and pH with video and imagery inputs.

Plant Ecology Beyond Land has developed a camera system for monitoring cultivated seaweed. Credit: PEBL

Christian Berger is CEO. He described a seaweed farm as essentially a series of buoys in the water. PEBL is working on systems to monitor how storms and other natural phenomena impact them. The key elements are a waterproof camera with filters to optimise contrast and software that identifies problems with the farm.

Monitoring the farms both above and below water, image analysis tracks certain rope positions and moorings, indicating potential issues within the farm.

Storm damage can be a danger to animals – particularly marine mammals such as dolphins and porpoises. If moorings are loosened, then the ropes may entangle these animals. PEBL’s work on ecological monitoring is funded by the World Wildlife Fund.

The equipment also monitors the cultivated seaweed, analysing growth rates and detecting disease. Regular scanning enables calculation of the total biomass in the facility. This is useful to understand growth factors but can also help monitor how much biomass is lost to the ocean due to storm damage. Understanding the fate of seaweed may become another important sustainability factor for aquaculture, as the carbon may be effectively sequestered in the deep ocean.

Berger said that adding seaweed within salmon farms would serve to naturally remove excess nitrates, providing more oxygen. For a truly sustainable future, integrating seaweed cultivation and fish farms may be a win-win situation for the aquaculture industry.

Image: Konstantin Zibert/shutterstock.com

09 November 2021

Image: Konstantin Zibert/shutterstock.com

09 November 2021

Outfield analyses aerial images of orchards using machine learning to predict crop yield

01 April 2021

Imec researchers on-board the parabolic flight. Credit Imec

12 January 2022

Mathias Bochow, GFZ Helmholtz Centre, Potsdam, is working on the Trace project to track marine plastic. Credit: Frank Schweikert, Aldebaran Marine Research & Broadcast (www.aldebaran.org/en/)

23 March 2022