Attention-enhanced and integrated deep learning approach for fishing vessel classification based on multiple features

Area of Science:

• Marine Biology
• Fisheries Science
• Artificial Intelligence

Background:

• Effective fisheries management relies on accurate monitoring of fishing vessels.
• Self-reported vessel data is often incomplete, hindering comprehensive management.
• Objective identification methods are crucial for improving fisheries monitoring.

Purpose of the Study:

• To develop an innovative deep learning model for classifying fishing vessel types using Automatic Identification System (AIS) data.
• To enhance the accuracy and completeness of fishing vessel monitoring for sustainable fisheries management.
• To classify five distinct fishing vessel types: gillnetter, hook and liner, trawler, fish carrier, and stow net vessel.

Main Methods:

• Preprocessing AIS data to ensure reliability and completeness of vessel trajectories.
• Constructing a multidimensional feature vector incorporating geometric, static, and dynamic vessel characteristics.
• Employing an ensemble model combining a 2D bidirectional long short-term memory network and a convolutional neural network with an attention mechanism.

Main Results:

• The integrated deep learning model achieved a classification accuracy of 91.90%.
• This accuracy surpasses that of individual classifiers, demonstrating superior performance.
• The method effectively differentiates behavioral patterns of various fishing vessel types.

Conclusions:

• The developed deep learning model offers a precise and objective method for classifying fishing vessels using AIS data.
• This approach significantly improves the performance of fishing vessel classification, aiding sustainable fisheries management.
• The model's remarkable performance indicates its potential for widespread adoption in fisheries monitoring systems.