Fisheries monitoring
Fish provide a valuable source of protein for human communities all over the world. To ensure fish stocks are sustained for future generations, regulatory bodies responsible for overseeing harvesting operations need reliable and comprehensive data on all species caught. This includes the targeted species and species incidentally caught (bycatch). Bycatch may include unmarketable fish, but also protected species. Regulators also need data that provide confidence fishers are compliant with regulations that are implemented to facilitate ecosystem sustainability outcomes (such as compliance with regulations that minimise interactions with protected species).
These data are collected through fisher dependent (e.g. logbook, catch data record) and independent sources. Historically, human observers on board vessels have been the primary source of independent at-sea fisheries data, but increasingly Electronic Monitoring (EM) programs are providing fishery independent video data for management purposes.
Electronic monitoring uses cameras, often in association with sensors and hardware, to record fishing activity. Data extracted from video imagery can be used to verify catch and monitor compliance.
One of the major challenges with EM programs is the need to cost-efficiently review the large amounts of video footage collected to generate the data used for fisheries management. MVT is using Artificial Intelligence/Machine Learning (AIML) technologies to realise several benefits: reduced management costs, expedited review time, and increased coverage for key target activities.
MVT staff have expertly extracted and labelled thousands of still images from EM videos which are then used to train AIML algorithms to correctly detect and classify objects of interest. We have developed cutting edge AIML software and tools that can assess the quality of video footage and automatically detect, count and book-mark fishing activity events. This enables efficient navigation to flagged events and automated filtering of fishing activities. A species identification component includes ‘confidence scores’, enabling auditors to assess and correct species identifications as needed. Our software also includes a reporting tool that summarises catch events and associated meta data for each video.
The ongoing development of CSIRO’s AIML technology will significantly contribute to advancing the quantity and quality of data collected and used for fisheries management and monitoring.
Watch how MVT can make a difference to fisheries management:
[Music plays and an animation image appears of a fishing boat on the surface of the ocean and fish swimming beneath in the water]
Global seafood sustainability in modern fisheries is essential.
[Animation image shows an inset set of balance scales and the image shows a dollar sign on one side and a fish on the other side of the scales]
It is also an expectation of our communities.
[Camera zooms in on the fishing boat and the image shows data lines radiating out from the ship and then the image shows rain and dollar signs falling around the ship]
But to achieve sustainability we need data. Unfortunately collecting fisheries data is expensive and at times dangerous.
[Animation image changes to show a computer screen showing the fishing vessel on the screen]
So, cameras are replacing humans as the collectors of fishing data on vessels.
[Animation image changes to show a male snoring in front of a computer screen which is displaying various types of fish]
Whilst more efficient, this leads to more hours of video than can be scrutinized.
[Animation image shows the camera panning along the computer screen showing the different types of fish and text appears: Whiting]
Advances in Artificial Intelligence and Machine Learning now allow these recordings to be analysed quickly and cheaply.
[Animation image shows an inset box displaying CSIRO data and text appears: AI Training, When – 05/09/19, Type – Whiting, Count – 50]
At CSIRO, Australia’s national science agency, we are training software to identify when a fish is caught, what type of fish is caught, and how many are caught.
[Animation image changes to show various types of fish on a bar graph]
This information helps scientists set sustainable annual catches, and it provides a better understanding of fishing impacts on the marine ecosystem.
[Animation image changes to show a fishing vessel on the surface of the ocean and fish swimming below]
For example, it will soon help manage unwanted bycatch of grenadier in the Sub-Antarctic.
[Camera zooms out on the fish and the vessel above]
With innovation and technology, we can solve the greatest challenges facing Australian fisheries and provide assurances of their sustainability now and into the future.
[Image changes to show the CSIRO logo on a light blue screen]