"Integrating Deep Learning into SAR Processing: From Reconstruction to Interpretation" by Francescopaolo Sica

Seminar by Francescopaolo Sica, University of the Bundeswehr Munich

Abstract

Synthetic Aperture Radar (SAR) has been a key technology in remote sensing for decades, offering robust Earth observation capabilities regardless of weather or lighting conditions. Yet, traditional SAR processing methods face persistent challenges, ranging from noise reduction and image reconstruction to high-level scene interpretation.

This seminar explores how recent advances in deep learning (DL) are addressing these challenges and reshaping the SAR processing pipeline. We will look at developments across three levels of abstraction: image formation, image enhancement, and semantic analysis.

Following a brief introduction to SAR fundamentals and image characteristics, the talk will highlight the limitations of conventional approaches and demonstrate how DL methods are being applied to improve both image quality and interpretability. Real-world examples and hands-on use cases will illustrate current trends, practical impacts, and the unique strengths of DL in tackling complex SAR problems.

This seminar will be held in english