Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/148471
Title: Vessel detection in Synthetic Aperture Radar images using Faster R-CNN models: Advanced monitoring techniques to improve fisheries management
Author: Carbó Mestre, Pol
Tutor: Rebrij, Romina  
Others: Ventura, Carles  
Abstract: Fisheries management has traditionally relied on catch quotas and manual reporting methods. However, the depletion of fish stocks has required the adoption of electronic reporting systems such as the Vessel Monitoring System (VMS) and Automatic Identification System (AIS), with its own limitations in preventing overexploitation. To address this, advancements in Synthetic Aperture Radar (SAR) satellite imagery analysis, coupled with Machine Learning (ML) techniques, have emerged as promising tools for monitoring fishing activities. These advancements are revolutionizing marine industry monitoring, closing the data gaps from traditional techniques, and enhancing transparency. This project focuses on vessel detection, employing computer vision methods. Convolutional Neural Networks (CNN), specifically Faster Region-Based CNN (Faster-RCNN), exhibit promising results with reduced detection time and computational costs. We evaluated the model's performance of these models and implemented various image pre-processing techniques to improve them. Furthermore, to demonstrate the potential of this approach in fisheries management, we tested the model using real-world Sentinel-1 images in a case study on Chilean fisheries and developed an interactive report presenting the results. The integration of SAR-based satellite imagery analysis and ML techniques holds significant promise for enhancing fisheries management. The evaluation of Faster-RCNN for vessel detection, along with the comparative analysis of pre-processing techniques, provides valuable insights into the effectiveness of this method. Furthermore, it also revealed some limitations of these techniques, underscoring the need for further advancements and emphasizing the reliance on combined approaches for effective fisheries management.
Keywords: machine learning
fisheries management
Synthetic Aperture Radar
Document type: info:eu-repo/semantics/masterThesis
Issue Date: 20-Jun-2023
Publication license: http://creativecommons.org/licenses/by-nc-sa/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
UOC_MemoriaFinal_pcarbomestre.pdfTFM report1,8 MBAdobe PDFThumbnail
View/Open
Share:
Export:
View statistics

This item is licensed under aCreative Commons License Creative Commons