Safer Grounds: A Study of Landmine Detection using UAV- and Ground-Based Multi-Modal Geophysics

Abstract

This dissertation addresses the urgent global crisis of landmines, unexploded ordnance (UXO), and explosive remnants of war (ERW) through the lens of multimodal geophysics. Chapter 1 sets the stage by highlighting the humanitarian imperative while underscoring the broader applicability of the developed methods and instruments for shallow critical zone exploration. Unlike conventional engineering-centric approaches, our geoscience-centered methodology offers promising avenues for effectively detecting and characterizing buried hazards. Chapter 2 meticulously examines various geophysical sensors, identifying limitations and proposing innovative solutions. Notably, TetraMag, a novel triaxial magnetic gradiometer, overcomes the deficiencies of single-sensor systems, demonstrating superior sensitivity to small-scale variations in the magnetic field. Chapter 3 delves into the intricate symmetries and invariants of the finite-difference magnetic gradient tensor (FDMGT), elucidating its pivotal role in precise target localization and parameter estimation within the shallow critical zone. The methodology outlined streamlines data processing and interpretation, laying a robust foundation for UAV-based detection systems. Chapter 4 introduces machine learning techniques, particularly convolutional neural networks (CNNs), as robust target detection and parameter estimation tools. By synergizing multiple geophysical modalities, these methods enhance our ability to discern subtle anomalies with high accuracy. Chapter 5 proposes a method to mitigate magnetic self-noise in UAV-mounted gradiometers, enhancing data fidelity and spatial coherence. This approach, applicable to various vehicle platforms, further extends the reach of our detection capabilities. In Chapter 6, we integrate and apply these methodologies to a real-world minefield scenario, successfully detecting and localizing buried targets. While acknowledging limitations such as payload constraints and computational demands, our findings underscore the versatility and robustness of the developed techniques. This dissertation addresses the pressing humanitarian challenge of landmine detection and advances the broader field of shallow critical zone geophysics. The methodologies and technologies presented here hold promise for diverse applications beyond military contexts, ranging from infrastructure mapping to hydrogeological studies.