Artificial and human intelligence used to tackle illegal fishing

By Fiona BrownAugust 21st, 2020

CSIRO is working with Microsoft and fisheries experts to harness robot and human-derived intelligence in the fight against illegal fishing.

Silently watching and listening in the waters off Indonesia and the Great Barrier Reef, robots lie in wait, patiently waiting for their targets to appear. While on land in Chile, information is being systematically drawn from enforcement officers to tackle clandestine behaviour. Two contrasting approaches with a similar goal – putting a stop to illegal, unregulated and unreported (IUU) fishing.  

They may sound sinister, but the robots are part of a project CSIRO is working on in collaboration with Microsoft 

We are combining robotic camera technology and underwater sound sensors (called hydrophones) to detect potential illegal fishing activities, such as fishing with explosives in Indonesia,  and improve our understanding about users of Australian marine reserves, so we can learn how to best manage reserves such as the Great Barrier Reef,” says Dr Chris Wilcox, Chief Research Scientist with CSIRO Oceans and Atmosphere.  

Utilising Microsoft’s machine learning tools and capabilities, we are ’training’ the robots to detect suspicious activity and send a real-time notification to law enforcement agencies.

In contrast, a complementary project in Chile is harnessing human-intelligence in the fight against illegal fishing. Working with enforcement officials from Chile’s national fisheries agency, CSIRO has partnered with colleagues from Cornell University and the Catholic University of Chile to develop a structured process for gathering expert knowledge and using it to estimate the level, structure and characteristics of illegal fishing. It is designed to complement existing approaches, providing a cost-effective, rapid and rigorous method to measure, monitor and inform solutions to reduce illegal fishing.  

The impact of illegal fishing

Illegal fishing is recognised as one of the main challenges to sustainable fisheries management and marine biodiversity conservation worldwide. Globally, approximately eight per cent of the world’s population depend on fishing for their livelihood. While Australia has some of the world’s most sustainably managed fisheries, unfortunately many of the world’s fisheries are in trouble; more than 80 per cent are either at full capacity or overexploited. Illegal fishing is a big contributor to this. An estimated 26 million tonnes of fish are caught through illegal fishing each year worldwide 

“Illegal fishing is the third most lucrative international crime behind weapons running and drug smuggling,” says Dr Wilcox. 

It affects about a third of the fish in the market and the livelihoods of 120 million people worldwide. So, it’s a major problem.” 

A fishing boat (left) and a blast explosion in the water (right)

Blast fishing is an illegal fishing method used to harvest and stun fish.

Illegal fishing activities have environmental, economic and social impacts, with developing nations whose people rely on fish as their primary source of protein and income suffering the greatest toll. Environmental impacts include inhibiting the recovery and sustainable management of fisheries stocks, increasing species endangerment, and destruction of marine habitats. Economically, illegal fishing is estimated to cost legal fisheries up to $23 billion annually. Illegal fishing can also be linked to a variety of other crimes including labour exploitation, human rights abuses and trafficking of illicit goods. 

Slipping through the net

Its impacts are diverse and far-reaching yet monitoring illegal fishing activity can be extremely difficult. Existing methods include ship tracking systems, fish catch data, and inspections at sea or in port. However, these approaches can be prohibitively expensive, particularly for developing countries. For example, Indonesia has roughly half a million fishing vessels, making inspecting individual vessels nearly impossible and tracking using technologies like GPS very challengingcurrent capacity allows authorities to track only 5,000 vessels on a daily basis. Similarly, with one of the world’s largest fishing zones, the area is challenging to patrol effectively and the cost of commercially available satellite imagery is unaffordable. 

Rise of the machines

A blue cylinder with what looks like a black bottle on top. A hydrophone.

A hydrophone is an underwater listening device used to detect illegal fishing.

To tackle the problem of constantly monitoring large areas of the ocean in real-time, CSIRO scientists are deploying sophisticated technology, that utilises machine learning, in the field to speed up the process.  They are using high resolution robotic cameras to capture data on the type of boat, boat features, boat travel speed, idling, and activities associated with people on the boats, e.g. fishing and diving. They are also using hydrophones, or underwater listening devices, which can record sounds from vessel engines, air compressors, winches or the detonation of explosives from tens of kilometres away and up to 30 meters below the surface.

The team are working with Microsoft to develop new cost-effective data processing tools and algorithms to turn the collected image and sound data into usable information. The algorithms train the deployed units to recognise images and audio patterns of likely illegal fishing activities. Applying Microsoft’s machine learning technology is enabling the large sets of data to be processed in real time, meaning that illegal fishing activities can be detected quickly and efficiently. Once detected the machines can send an immediate alert to law enforcement agencies that suspicious fishing behaviour is occurring, increasing the speed at which an investigation can be launched and the likelihood that illegal activity will be caught.  

Ask those in the know

Fisheries managers, enforcement officers and government officials hold a wealth of informal information about illegal fishing activity in the areas they manage. However, this knowledge can be difficult to utilise due to issues with bias, transparency and fragmentation of information.  

“Working with enforcement officials in Chile’s national fisheries agency, we conducted online surveys and used statistical models to develop a framework agencies can use to monitor illegal fishing activity, discover leverage points to target investments in enforcement operations, and evaluate the impact of their interventions,” says Dr Wilcox. 

An enforcement officer, in a black suit and helmet, pictured at what looks like a market stand.

Enforcement officers hold a wealth of informal information about illegal fishing activities. Image credit: Shutterstock

The team used a structured process, that accounts for potential biases (such as experience and differences among individual officers) and reduces fragmentation of knowledge, enabling fisheries management agencies to formalise their institutional knowledge.   

Given that current approaches to measuring illegal fishing can be resource intensive and sometimes controversial, estimating illegal activity directly from fisheries enforcement officers is a complementary approach that provides a cost-effective, rapid, and rigorous method to measure, monitor, and inform solutions to reduce illegal fishing,” says Dr Wilcox. 

The right time and the right price

Gathering intelligence, be it ‘artificial’ from trained robots or from human experts, in a cost effective and efficient way, and developing methods to convert the resultant data into usable information about illegal fishing activities is helping fisheries managers get the information they need at the right time and the right price. This information will give them the best chance to effectively combat the issue of illegal fishing.  

“Ultimately, if we reduce the amount of illegal fishing, there will be wide ranging economic, environmental and social benefits for communities, consumers and businesses,” says Dr Wilcox. 

0 comments

Leave a Reply

Your email address will not be published. Required fields are marked *