Illegal fishing detection using machine learning – Automation for the sake of private firm profits or genuine benefit in tracking dark vessels?

Illegal fishing detection technologies that use artificial intelligence (AI) and machine learning (ml) employ a variety of tools to detect vessel behaviour and determine whether a vessel is illegally fishing based on location, speed, size, Geo-location alerts, and image object detection software to generate threat assessment picture. The data that has been auto-tuned by MLS (machine learning software) is then sent to law enforcement to investigate dark vessel activities. MLS is based on the fact that each vessel has a unique id based on vessel identifiers and digital sensors such as AIS, Radar, and GSM software sharing data such as a vessel's location, speed, and heading.

IUU Fishing

What is illegal fishing ?

The United Nations’ Food and Agriculture Organisation’s (FAO) definition of illegal, unreported, and unregulated fishing (IPOA on IUU Fishing, 2001) is too narrow in its focus to account for all types of violations in conventional marine fisheries, that could range from minor infractions to serious crimes committed at sea. With the transition of maritime law enforcement from traditional approaches using patrol boats and aircraft to digital tracking of fishing vessels, illegal fishing detection tools such as artificial intelligence and machine learning are receiving more attention from private for profit firms and satellite tracking companies selling machine learning software to fisheries and maritime institutions in developing countries. We can use machine learning software to predict whether a vessel is engaging in illegal fishing by automating video and satellite location feeds.

FAO definition of illegal fishing (FAO 2001)
Figure 1: Limited scope of the definition of IUU fishing as illustrated in the “FAO International Plan of Action on IUU Fishing, 2001 document”
  • Illegal fishing – (Unauthorized fishing, fishing without license, misreporting, under-reporting, fishing in marine reserves, fishing using explosives and chemicals)
  • Unreported catches (Reporting high value species as low value fish to avoid quota accounting, taxes, high-grading, discards, etc.)
  • Unregulated fishing (Unauthorized fishing on the high seas by vessels having no nationality or by vessels not a party to the relevant RFMO)

Why has IUU Fishing not received more attention until recently ?

  1. Unreported and unregulated fishing have garnered less attention than illegal fishing the threat receiving most focus in research and media. From the perspective of a coastal state that is a party to UNCLOS, the challenge to sovereignty posed by “illegal fishing” is more severe than unreported and unregulated fishing activities.
  2. From a taxes and catch-quantification standpoint, underdeveloped countries are more at risk from unreported fishing.
  3. From the perspective of flag state compliance and the breach of international agreements to which a coastal state is a signatory, unregulated fishing creates dangers.
  4. It’s important to remember that the marine domain is home to a wide range of criminal activities, and that each nation places different priorities for the containment of maritime threats. (Piracy, for instance, may be seen as a greater danger in Somalia than illicit fishing or narcotics trafficking)
  5. The extent of the IUU risk varies from nation to country, and some threats receive a greater share of surveillance resources. The location, size of the coastline, and Human Development Index (HDI) of a country all play a role in the level of priority given to certain marine risks. Although more than seven separate hazards might be identified using surveillance technologies, most cash-strapped underdeveloped countries can only dedicate patrol resources for one or two major threats.

Within the maritime EEZ limits, the types of THREATS and associated RISK vary by country and location (IUU Bubble)

  • Marine Terrorism
  • Narcotics trafficking
  • Human trafficking
  • Contraband smuggling
  • IUU fishing (is usually an “outlier” MDA control task)
IUU fishing baloon
Figure 3: Illegal fishing treated as a low priority security risk in the maritime domain. Size of the bubble for different maritime threats varies by each country.

illegal fishing detection using machine learning

Illegal fishing vessels are identified or spotted using a variety of data analytics technologies, including vessel location (AIS, VMS), satellite imagery (VIIRS, SAR), and coastal radars, which are increasingly being supplemented by newer and less expensive tools like drones and UAV.

Coastal radars, maritime patrol aircraft, and sea-based boardings were traditional surveillance tools. With the emergence of new affordable technologies over the last decade, the use of AIS and satellite imagery has become more cost-effective, and more machine learning tools, and software companies have emerged to design algorithms and machine learning software (MLS) tools to detect illegal fishing at sea.

Using Radio Frequency Analytics to identify illegal fishing vessels (HawkEye 360)

For starters, most countries do not have a legal mandate under fisheries and other maritime laws to use digital data collection systems that use artificial intelligence (AI) systems to send alerts to centralised maritime co-ordination Centres (MCC) or Fisheries Monitoring Centers (FMC) when suspicious and dark vessels are detected using VMS or AIS signals of licenced vessels and vessels declaring innocent passage through EEZ under UNCLOS statutes.

Second, there are several critical gaps in how data collected by private for profits firms and ENGOs providing AI and MLS technologies are used in relation to maritime security and “General Data Protection Rules” violations when such data is used to develop new software without the consent of third countries whose content is used in development of the new technologies and improving machine learning behaviour. I’ve written about it here.

Traditionally, machine learning (ML) is used to monitor compliance with by-catch, TED regulations, and other electronic monitoring systems that deploy onboard cameras on fishing vessels to replace human observers onboard vessels at sea. Such electronic camera systems monitor compliance by measuring the size of fish, the retention or discarding of non-quota or protected species, the length of the fishing vessel, the type of fishing activity, and so on.

Third, Artificial intelligence-based MLS systems can improve forecasting of actual catches and size of fish retained because footage is available for the entire trip for more detailed audits when compliance lapses are suspected. Some of the satellite surveillance technologies used in illegal fishing detection include Synthetic Aperture Radar (SAR) imagery, Radio Frequency analytics (RF) imagery, RADARSAT satellite imagery,

Detecting illegal vessels using Synthetic Aperture Radar (SAR) imagery (ICEYE)

Some of the current gaps in the AI-based MLS fisheries industry might be related to transparency, data collection rules, data storage, and ownership issues, which have never been formally regulated, even in the European Union or developing countries.

Detecting illegal fishing vessels using SAR imagery (MDA)
Use of RADARSAT satellite imagery to tackle illegal fishing

Suggested reading materials

Fernandes-Salvador, J.A., Oanta, G.A., Olivert-Amado, A., Goienetxea, I., Ibaibarriaga, L., Aranda, M., Cuende, E., Foti, G., Olabarrieta, I., Murua, J., Prellezo, R., Iñarra, B., Quincoces, I., Caballero, A., Sobrino- Heredia, J. M, (2022) Artificial Intelligence and the fisheries sector, European Parliament, Policy Department for Structural and Cohesion Policies, Brussels, Research for PECH Committee, 104 pages. 

Longépé, N., Hajduch, G., Ardianto, R., de Joux, R., Nhunfat, B., Marzuki, M. I., … & Gaspar, P. (2018). Completing fishing monitoring with spaceborne Vessel Detection System (VDS) and Automatic Identification System (AIS) to assess illegal fishing in IndonesiaMarine pollution bulletin131, 33-39.

M. Rizwan Khokher, L. Richard Little, Geoffrey N. Tuck, Daniel V. Smith, Maoying Qiao, Carlie Devine, Helen O’Neill, John J. Pogonoski, Rhys Arangio, and Dadong Wang. Early lessons in deploying cameras and artificial intelligence technology for fisheries catch monitoring: where machine learning meets commercial fishing. Canadian Journal of Fisheries and Aquatic Sciences79(2): 257-266.

May Petry, L., Soares, A., Bogorny, V., Brandoli, B., & Matwin, S. (2020, May). Challenges in vessel behavior and anomaly detection: From classical machine learning to deep learning. In Canadian Conference on Artificial Intelligence(pp. 401-407). Springer, Cham.

Leave a comment