engleski

Otkrivanje mina s malom količinom metala zasnovano na elektromagnetskom induktivnom modelu

Antipersonnel landmines are a global safety hazard that indiscriminately affects both soldiers and civilians, causing severe injuries and deaths years after the cessation of conflicts. This thesis presents an attempt to improve the existing methods of close-in detection of landmines, particularly those with low metallic content, using electromagnetic induction (EMI) sensing. A key assumption is that such improvement could be obtained by shifting from an existing concept of metal detection to a concept of metallic target characterization. This could ultimately lead to safer, faster and cheaper humanitarian demining. The landmine problem and the concept of humanitarian mine action are introduced first. State-of-the art methods of close-in landmine detection are briefly described, with an emphasis on metal detection as a key sensing modality. An analytical forward model of a scanning EMI sensor for metallic target characterization, based on the induced magnetic dipole mode, is devised from the governing EMI equations. Several representative search head designs are selected, which will be used throughout the thesis for conceptual tests. Three different static inversion methods for the estimation of induced dipole model parameters from EMI and sensor positional data are analyzed ; the method based on nonlinear least squares (NLS) approach, the quasi-numerical HAP method featuring the modified NSMS model, and the metaheuristic method. Inversion performance of each method is evaluated on a synthetic data set, for selected metallic targets and representative search head geometries. A novel approach to tracking of the search head's pose based on the measurement of its own primary magnetic field and the application of the extended Kalman filter is presented. Following a similar methodology based on Bayesian filtering, a new concept of dynamic dipole inversion that avoids the need for sensor positional data is proposed. The effects of background interferences on a response of an EMI sensor for use in humanitarian demining and the methods for their compensation are also analyzed. A novel concept for robust estimation of target's shape, based on temporal information stored in the eigenvalues of the estimated magnetic polarizability tensor, is introduced. Finally, all the presented modeling, inversion, tracking and compensation approaches are experimentally validated on several sensing platforms in a laboratory and field environment.

metal detection, metal characterization, electromagnetic induction, induced magnetic dipole, magnetic polarizability tensor, nonlinear least squares, magnetic tracking

nije evidentirano