Improved resource control of fisheries using deep learning

State of the Art, Knowledge Needs, and Project Goals

• The main objective is to enhance resource control in the fishing industry to ensure the sustainable utilization of marine resources, in line with the UN Sustainable Development Goal #14.
• Current manual reporting systems are prone to misreporting, which can lead to unjustified economic gains.
• The project will utilize Automatic Identification System (AIS) data and deep learning to predict which ports fishing vessels will land their catch.
• The goal is to develop a predictive system that can assist fishery inspectors in being present during unloading, thereby minimizing the risk of manipulation of catch data.

Research Questions, Hypotheses, Theoretical Approach, and Methodology

• The project will investigate whether it is possible to develop a deep learning model that can predict port arrivals with high precision based on historical AIS data and other relevant information.
• It will also explore whether historical catch reports can be analyzed to detect anomalies in catch composition, which may indicate misreporting.
• Another aspect of the project will focus on using machine learning to detect fishing activity and gear usage based on live AIS data.

Innovation and Ambition

• The ambition is to develop an integrated model that links AIS tracking of fishing vessels with destination prediction, fishing activity detection, and identification of potential manipulation in catch reports.
• The project will employ advanced machine learning methods to improve the efficiency of resource control.

Potential Impact of the Proposed Research

• The project will directly benefit Råfisklaget by making inspection work more efficient and contributing to the sustainable management of fishery resources.
• The results will have international relevance and can be applied in other contexts such as port operations and search and rescue (SAR).
• The project will also help strengthen the research environment at UiT in the analysis of multivariate time series using deep learning.

Communication and Utilization Measures

• The results will be disseminated through national and international conferences, scientific journals, and open data-sharing platforms.
• The project will also involve master’s students and explore opportunities for innovation and industrialization of the research.

Implementation

• The project is led by Associate Professor Helge Fredriksen at UiT, Department of Computer Science, with contributions from other researchers and master’s students.
• Råfisklaget will contribute expertise in the fishing industry and facilitate data access.