Do you know the story behind the National Day Seafood Feast?
Did you enjoy the seafood feast on October 1st?
The happy National Day holiday is about to pass. Have you enjoyed a seafood feast? A sumptuous seafood feast always brings us a sense of satisfaction. However, have you ever thought about what it would be like if one day we could no longer taste delicious seafood?
In fact, behind our ability to enjoy marine delicacies year after year are the efforts of some people to combat illegal fishing, a behavior that seriously damages the fishery ecology, affects fishery production, and hinders the sustainable development of capture fisheries.
Pictured: Fisheries observers track fish caught on fishing vessels.
Photo credit: Mary Catherine O'Connor.
According to statistics from the Western and Central Pacific Fisheries Commission, fisheries losses due to illegal fishing are about $1.5 billion each year. In order to avoid these huge losses, fishery observers need to work almost 24/7. They cruise the Pacific Ocean on a fishing boat hundreds of miles from the coastline. Even though the work is so hard, there is no rich reward, especially when observers have to track the source of seafood caught by fishing boats, they not only have to be disgusted by fishermen, but also encounter obstacles brought to them by fishermen. This source tracking is mainly used for bycatch to avoid possible illegal fishing.
So is there any way to solve or improve this problem?
The relevant people thought of deploying artificial intelligence seafarers to combat illegal fishing. But what is the reason behind this?
It is estimated that about 20% of the world's seafood products come from illegal fishing. These illegal activities mainly occur in licensed fishing vessels, not unlicensed fishing vessels as we imagine.
Therefore, experts from The Nature Conservancy (TNC) are actively promoting the use of AI-equipped video monitors on commercial fishing vessels. The software used in the monitors has machine learning capabilities and can quickly determine the species of each fish based on its size, shape and color. Fishery regulators can also rely on monitors and corresponding software systems to inspect fishing vessels to determine whether their fishing process is legal. This is a new artificial intelligence (AI) monitoring system , which can be regarded as a "face recognition system" for fish . This system can help fishery observers in their work, help maintain the population of fish and maintain the balance of the entire marine ecology. And it also allows people to gradually free themselves from the heavy and hard work of fishery observers.
Although AI technology has been widely used in more and more industries, it encountered a lot of resistance when it was first applied. Mark Zimring, director of TNC's Indian and Pacific tuna project, said that this technology was doubted by most people in the industry in its early stages. The opposition was within Mark's expectations. After all, AI image recognition technology was not mature at that time. It was very difficult to identify objects with a white screen background in a controlled static environment or to identify objects with a stable speed on a conveyor belt.
But in the end, they still made this idea a reality. During the test, they made salt stains on the lens, used different lighting intensities to simulate day and night, and placed fish in various positions within the monitoring range for identification, trying their best to reproduce the situation that might occur on a real fishing boat.
Compared to the past when fishery observers were solely responsible for reporting illegal activities to international regulatory agencies, the emergence of AI monitoring modules is a big improvement.
Does the “fish world” also have facial recognition?
These cameras (electronic monitors) placed on the deck can replace fishery observers. The cameras are connected to the ship's communication system to provide information to the regulatory authorities. If someone tries to tamper with the camera content or cover the lens, the alarm system will be triggered.
After the catch, it typically takes hundreds of hours to review the images sent back by the cameras, so TNC set out to develop an AI algorithm similar to the "facial recognition" of fish .
In order to complete the recognition algorithm, TNC launched a competition on Kaggle (a platform for developers and data scientists to hold machine learning competitions, host databases, write and share codes). In the end, nearly 2,300 teams participated in the competition and submitted codes, making it one of the most participated competitions on Kaggle. The five winning teams shared a prize of $150,000. Subsequently, TNA tested the effectiveness of the code with Satlink, a manufacturer of electronic monitors.
“Our goal was to provide reference examples and web services for existing electronic surveillance systems to achieve the identification capabilities we needed,” said Matt Merrifield, leader of TNC’s Geographic Information Systems (GIS) team. “This meant improving existing resources rather than designing entirely new workflows and algorithms.”
Chris Rodley, CEO of Snap Information Technologies, a New Zealand company that helped TNC improve its AI algorithms, said the software system used by TNC can autonomously learn from its incorrect judgments and gradually improve its accuracy.
The machine learning algorithm uses existing electronic surveillance images to learn the physical characteristics of different types of fish under different lighting conditions, and professional fishery observers then review the software's findings.
Rodley and most others are optimistic about the system, which they expect will reduce the time it takes to review electronic surveillance footage from more than 40 hours to just a few hours. This could significantly reduce the time fishery observers spend on video review, freeing them up to do other work.
AI brings a bright future for sustainable fisheries
The TNC’s goal is to increase transparency in the fishing industry and make fleets pay for their illegal behavior. If AI software can indeed accurately identify fish species and effectively reduce labor costs, its application range may be further expanded.
Large fishing companies can use AI software to improve transparency in their industry and use this as a competitive advantage; retailers can also require their suppliers to use electronic monitoring systems to ensure the legality of their goods.
Francisco Blaha, a professional consultant to governments and regulators in the fisheries field, said that AI does have great potential, but there may still be some obstacles in its practical application. He believes that technology suppliers and regulators should work together to develop a legal framework so that video evidence produced by AI can be legally recognized, which should include the validity of the evidence and the acceptable range of error.
AI electronic monitoring could ease the burden of fishery observers
Amount of manual work on board.
Image credit: Mary Catherine O'Connor
Blaha said that for AI to really work in fisheries, a lot of resources and optimization will be needed, especially for the unstable light intensity and inevitable shaking on fishing vessels.
In addition, since the technology is not promoted by the government and regulatory authorities, but by TNC and electronic monitoring suppliers, there is no exact time for it to be put into use. However, Blaha still believes that it is worth the effort. After all, fish are an important source of protein worldwide, and the entire fishery needs greater transparency and accountability.
Zimring from TNC believes that the number of fishery observers is far from enough compared to the number of fishing vessels, and AI facial recognition software will play an irreplaceable role in the future and may ultimately prevent illegal fishing activities.
* Recommended Reading *
Click [Read original text] to view more exciting content