Congratulations to our ESR 11 from MENELAOSแดบแต, Mobina Keymasi, who successfully defended her PhD thesis on the topic of “Goal Directed Compressive Radar Imaging ” on November 20, 2024 at the University POLITEHNICA of Bucharest under the supervision of Prof. Mihai Datcu.
The domain of remote sensing (RS), especially Synthetic Aperture Radar (SAR) imaging, has undergone a major shift due to the incorporation of artificial intelligence (AI) and machine learning (ML) methods. These technologies are now crucial for managing the increasingly intricate and voluminous datasets produced by modern Earth observation (EO) missions. However, this rapid expansion of data brings forth new difficulties, particularly related to data handling, storage, and real-time analysis, especially when working with high-dimensional SAR and multispectral data.
In response to these issues, this thesis introduces innovative approaches utilizing compres-sive sensing (CS) for both SAR and optical data. Compressive sensing is a mathematical method that facilitates efficient data acquisition by minimizing the number of measure-ments required for accurate image reconstruction. This results in faster data processing while maintaining image quality, making CS particularly advantageous for SAR applica-tions in environmental monitoring, where both accuracy and timeliness of information are crucial.
The core techniques developed in this work include the creation of semantic frameworks for detecting ships in SAR imagery, the merging of SAR with multispectral datasets, and the application of convolutional neural networks (CNNs) for advanced feature extraction. These methods are employed in complex environments, such as the Danube Delta, where monitoring environmental changes over time necessitates the integration of diverse remote sensing data sources.
Moreover, this research introduces a hybrid framework that integrates compressive sensing (CS) with machine learning (ML) algorithms to enhance both data acquisition and processing efficiency. The proposed method is validated through several case studies, demonstrating the combination of SAR and multispectral data for monitoring environmental changes and identifying specific objects, such as ships.
The findings reveal notable improvements in both accuracy and efficiency in remote sensing tasks, particularly by addressing the complexity associated with large-scale Earth observation (EO) data.
The study highlights the significance of data fusion and scalable AI solutions in over-coming the challenges faced by modern EO systems. By combining CS with AI/ML methodologies, the research offers a comprehensive approach to improving the resolu-tion, accuracy, and processing performance of remote sensing data. This synergy not only enhances the monitoring of Earthโs natural resources but also lays the groundwork for future innovations in radar-based EO technologies.
To conclude, this thesis contributes to the advancement of remote sensing by proposing innovative strategies that address key issues in data acquisition and processing. These strategies enable more precise and efficient monitoring of both environmental and urban changes. The research emphasizes the potential of AI and CS to push the frontiers of SAR-based Earth observation, providing robust and scalable solutions for real-time environmental monitoring and resource management.
Recent Comments