Esposito, Giuseppe (2022) GNC and imaging approaches for mini-UAV based radar. [Tesi di dottorato]

Anteprima

Testo esposito_giuseppe_35.pdf Download (6MB) | Anteprima

Abstract

Radar imaging from Unmanned Aerial Vehicles (UAVs) is a dynamic research topic attracting huge interest due to its practical fallouts. This thesis deals with mini-UAV based radar imaging and faces several open issues related to surface and subsurface imaging from microwave data collected by a radar system mounted on board a mini UAV. Initially, a brief review of the radar imaging technologies dealing with the state of the art systems, the required hardware components and the applicative fields is presented. Thereafter, the main open issues are introduced. These are: 1) the need for effective filtering strategies to reduce the clutter; 2) the need for accurate UAV positions estimation during the data acquisition; 3) the need for obtaining from radar data accurate and high resolution images of the observed scene, while keeping acceptable the computing times and the computer resources in terms of memory space and processor velocity. In this regard, this thesis reviews the data filtering procedures commonly adopted in Time Domain (TD) Ground Penetrating Radar (GPR) literature, it analyses different possible solutions for obtaining accurate UAV positioning information, and proposes two different strategies to account for the UAV positioning estimates. Both the strategies, referred to as Strategy ‘A’ and Strategy ‘B’ respectively, use filtering strategies operating in time domain, the Fast Fourier Transform to transform data in the frequency domain, and a microwave tomography approach to process data. This latter addresses the imaging as a linearized inverse scattering problem. As said, the strategies differ from each other in the way they account for UAV positioning information. Strategy 'A' implements a Motion Compensation (MoCo) step that accounts for the UAV motion deviations with respect to nominal trajectory and exploits the UAV positioning information to realign the radar data and ensure their uniform spacing along a straight trajectory. This strategy also involves the use of the 'Shift and Zoom' procedure to improve the computational efficiency of the microwave tomographic (MWT) approach adopted to face the radar imaging problem. The Shift and Zoom consists in dividing the observation domain and the investigation one into smaller and partially overlapping subdomains in which the MWT approach is used to obtain tomographic reconstructions that are then combined together to obtain the reconstruction of the entire domain under test. It should be noted that having data uniformly spaced along a straight trajectory at constant altitude and using the 'Shift and Zoom' approach allow a particularly efficient implementation of MWT because it is sufficient to calculate the scattering operator, i.e. the mathematical operator linking the data and the unknown, once and for reduced dimensions of the observation domain and the investigation one. Strategy 'B', conversely, uses the platform positioning data directly into the reconstruction step, i.e. in the implementation of the MWT approach. This avoids possible data alterations due to the resampling and the interpolation required by the MoCo step of Strategy ‘A’. In addition, Strategy ‘B’ is suitable for arbitrary flight geometries, i.e. not only for straight trajectories. It is worth noting that both strategies ‘A’ and ‘B’ have been used to process experimental data. A further, and maybe most important, contribute of this thesis regards the design of MWT approaches to face the imaging of surface or subsurface targets from radar data collected by using a mini UAV as observation platform. It is worth pointing out that the radar imaging problem has been addressed in the case of two-dimensional geometry considering both the vertical imaging plane, i.e. the plane defined by the flight path and the pointing direction of the transmitting and receiving antennas, and the horizontal imaging plane, i.e. the plane at constant altitude. Accordingly, after a brief review of the basic concepts regarding MWT, the approaches developed during the PhD activity are presented. These MWT approaches differ from each other for the scattering model adopted to describe signal propagation while using the same mathematical tools to solve the inverse scattering problem and allowing the exploitation of some figures of merit for analysing the achievable spatial resolution limits. In this frame, a strategy called MIA (Multiline Imaging Approach) is also proposed. MIA considers the imaging in the vertical plane and exploits the radar data collected on one or more measurement lines to reconstruct 2D domains (slices), which are then interpolated to provide a pseudo-3D representation of the investigated volume. The computational burden of MIA is significantly reduced compared to that required by a full 3D approach. This thesis also envisages the use of UAV radar systems for inspections of surface and subsurface scenarios. Specifically, two high frequency radar systems are considered, referred to as System HI and System HII, and two low frequency radar systems, System LI and System LII. These systems were used in various measurement campaigns concerning objects placed on the surface (Systems HI and HII) and buried (Systems LI and LII) and the acquired radar data were used for the experimental validation of the proposed strategies and of the designed MWT approaches. Specifically, Strategy 'A' was successfully tested on radar data acquired through the System HI and allowed the comparison of radar imaging performance when using standalone and differential Global Navigation Satellite System (GNSS) positioning information. In this case the Strategy ‘A’ was implemented by using the MWT approach based on the vertical imaging model for free-space propagation. Strategy 'B' was tested both on high-frequency radar data acquired via System HI, and on low-frequency radar data acquired via System LI. In the first case, Strategy 'B' was implemented by exploiting a MWT approach formulating the imaging in the horizontal plane and considering the electromagnetic propagation model in free space. Moreover, the reconstruction capabilities of the system were analysed showing the effect of the radar parameters, i.e. the flight altitude and the spatial offset between antennas and targets, on the resolution limits, and the consistency of the results with the theoretical resolution limits was demonstrated. Results also demonstrated that when a target is observed off-nadir, a slightly curved trajectory can help distinguishing the real target from the ghost target related to left-right ambiguity. In the second case, Strategy 'B' was implemented by using a MWT approach formulating the imaging in the vertical plane and considering the presence of the air-ground interface. Specifically, among the vertical imaging models proposed for the subsurface propagation, the Equivalent Permittivity (EP) model was used to describe the electromagnetic propagation in a non-homogeneous medium. The obtained results validate the ability to identify and locate buried objects, specifically a metal plate placed about 30 cm below the air-ground interface. The MIA strategy was tested on the data acquired via System HII at the archaeological park of Paestum and Velia and the obtained results demonstrated good ability to focus and localize the targets in the investigated scene. Finally, this thesis reports some preliminary results referred to the System LII, which was designed and realized during the abroad PhD period. The System LII was realized by using a Vector Network Analyzer (VNA), was calibrated by performing reference measurements, and was tested by inspecting the internal structure of a concrete wall. The obtained experimental results were compared with results obtained from numerical simulations.

Downloads

Actions (login required)

Modifica documento