FAD-SAR: A Novel Fishing Activity Detection System via Synthetic Aperture Radar Images Based on Deep Learning Method

TL;DR

This work tackles illegal, unreported, and unregulated (IUU) fishing detection using Synthetic Aperture Radar (SAR) imagery by proposing a deep learning–based fishing-activity detection system evaluated on the xView3 dataset. It assesses six classical detectors (SSD, RetinaNet, FSAF, FCOS, Faster R-CNN, Cascade R-CNN) and implements data processing and enhancement techniques, notably Online Hard Example Mining (OHEM) to train Faster R-CNN, achieving an Avg-F1 improvement from $0.212$ to $0.216$. The study demonstrates that two-stage detectors generally outperform one-stage ones on xView3, and shows that carefully designed data processing (800×800 patches, VV/VH channels with average embedding) and augmentation strategies can stabilize SAR-based detection of fishing activity. The findings highlight the potential of DL-based SAR analysis for automated IUU surveillance and provide a concrete performance baseline for SAR-driven fishing-activity detection. $Avg ext{-}F1$ improvements and model comparisons offer actionable guidance for deploying SAR–DL systems in maritime monitoring contexts.

Abstract

Illegal, unreported, and unregulated (IUU) fishing activities seriously affect various aspects of human life. However, traditional methods for detecting and monitoring IUU fishing activities at sea have limitations. Although synthetic aperture radar (SAR) can complement existing vessel detection systems, extracting useful information from SAR images using traditional methods remains a challenge, especially in IUU fishing. This paper proposes a deep learning based fishing activity detection system, which is implemented on the xView3 dataset using six classical object detection models: SSD, RetinaNet, FSAF, FCOS, Faster R-CNN, and Cascade R-CNN. In addition, this work employs different enhancement techniques to improve the performance of the Faster R-CNN model. The experimental results demonstrate that training the Faster R-CNN model using the Online Hard Example Mining (OHEM) strategy increases the Avg-F1 value from 0.212 to 0.216.

Figures (2)

• Figure 1: The overall architecture of the proposed fishing activity detection system, mainly consists of the training of deep learning models, and the employment of enhancement techniques to improve the model performance.
• Figure 2: Example results of different models for fishing activity detection.