The subject of this thesis is the application of statistical filtering techniques to passive radar. The objective of this thesis is to adapt Gaussian sum filtering and particle filtering methods to the detection and/or tracking in a multi-target context. Highly manoeuvring targets, at potentially very low signal-to-noise ratios, will be of particular interest. As an application, FM radio and terrestrial digital video broadcasting (DVB-T) will be exploited as illuminators of opportunity by the passive localization system. First, this thesis recapitulates the state-of-the-art in the domain of passive radar, statistical filtering and conventional radar tracking approaches based on the thresholded data. Second, this thesis explores the benefits of particle filtering in passive radar. With an appropriate modeling of the problem of target tracking as a non-linear dynamical system, we show how particle filtering, fed with the noisy unthresholded matched filter outputs, outperforms conventional tracking approaches. An extension to the multi-target case is also treated. The essential ingredient of the proposed algorithm is the inbuilt synchronization system of the correlator sampling instants (and potentially also of the correlation frequency), which allows the particle filter to compensate the dynamics of the targets automatically. Third, we present a new multi-target detection/tracking system, based on Bayesian filtering, using the track-before-detect methodology. This system is implemented with an approximation based on Gaussian sum filtering or an approximation based on particle filtering. We also propose a successive interference cancellation technique, which allows to handle the presence of large sidelobes. Simulations using FM radio confirmed the potential of the proposed detection/tracking system.
