Thesis

Advanced maritime target recognition from SAR images exploiting target’s micro-motions and AI

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2026
Thesis identifier
  • T18019
Person Identifier (Local)
  • 202060639
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • The work presented in this thesis is an exploration of advanced techniques for maritime target recognition using Synthetic Aperture Radar (SAR) data, focusing on innovative approaches to target classification, co-registration, and micro-motion analysis, in the field of maritime surveillance. SAR technology offers unique advantages in maritime environments, including the ability to operate under diverse weather and lighting conditions. This research aims to exploit these capabilities to address the challenges of detecting and classifying maritime targets, particularly in scenarios involving moving or vibrating ships due to their engine. The literature review establishes the foundation by discussing SAR geometry, image formation processes, and acquisition modes, providing a comprehensive understanding of the principles that underpin SAR data and imaging. It also highlights state-of-the-art advancements in co-registration of multitemporal SAR images and micro-motion analysis, offering insights into their applications in maritime environments. A novel cross-cross-correlation-based method is proposed for joint co-registration of rotated multitemporal SAR images, addressing challenges related to alignment and displacement field estimation. The effectiveness of the method is demonstrated through experimental evaluations, highlighting its potential for improving SAR image processing workflows for maritime surveillance. This research also introduces a novel framework for maritime target classification using Single Look Complex (SLC) SAR data. Key contributions include the extraction of spectral profile features and the integration of invariant features, such as Krawtchouk moments, which enhance the classification performance of the overall system. A combination of machine learning and deep learning techniques, including Recurrent Neural Networks (RNNs) and confidence-based ensemble methods, is employed to improve accuracy and resilience against noise. Furthermore, a second framework for detection, location, and classification of targets based on autocorrelation is proposed. By exploiting sub-aperture processing and physics-informed feature extraction, the framework enables robust identification of vibrating targets and discrimination from static objects and clutter. The results demonstrate that micro-motion signatures embedded in the autocorrelation domain can be effectively characterised and leveraged for target classification, improving sensitivity to subtle object dynamics. The findings presented in this thesis introduce novel methodologies for target recognition and categorisation in SAR-based maritime surveillance. By incorporating micromotion–aware features, the proposed approaches enhance the capability of SAR systems to detect, characterise, and interpret maritime activities. These advancements have direct relevance to military surveillance, environmental monitoring, and the detection of illegal or anomalous maritime behaviour.
Advisor / supervisor
  • Clemente, Carmine (Reader)
  • Ilioudis, Christos V.
  • Macdonald, Malcolm (Aerospace engineer)
Resource Type
DOI
Funder

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T18019 2026-05-21 Public Download