Imaging Algorithms For UWB Through-wall Radar

Sensing through obstacles such as walls, doors, and other visually opaque materials using microwave signals, is emerging as a powerful tool which supports a range of civilian and military applications. Through-the-wall radar imaging (TWR1) has recently sought out for surveillance and reconnaissance in urban environments, which can be employed to detect and locate survivors for the succors in search and rescue in natural disasters, such as earthquakes and avalanches.The critical problems which need to be solved in through-wall radar imaging algorithms are nonlinearity, ill-posedness, effects of walls and real-time. Aimed at the four keys, the systematic studies are carried out in this dissertation based on the simulations of through-wall scenario using finite difference time domain (FDTD), and the main achievements are as follows:Firstly, characterizations of wall effects on ultra-wideband (UWB) signals are studied. UWB signals, when propagating through walls, suffer signal delay, attenuation and distortion. The effects of frequency of signals, dielectric properties or thickness of walls and information of target on the received signals are discussed.Secondly, a real-time approach for solving the through-wall imaging is proposed based on support vector machine (SVM). The proposed technique avoids computation of the effect of the walls and deals with nonlinearity and ill-posedness inherent in the through-wall imaging problem successfully. Through-wall imaging problem is converted to the establishment and use of a mapping between backscattered data and the information of the target. Then the nonlinearity and the propagation effects caused by walls are included in the mapping that can be regressed after SVM training process. The hyper-parameters in SVM have regularization effects. Once the training phase is completed, the proposed technique only needs computational time in an order of seconds to predict. The feasibility and validity is tested and the results demonstrate that this approach can detect and position the targets behind the wall whether there is noise or not.Thirdly, a real-time method based on the relevance vector machine (RVM) for through-wall imaging problem is also presented. The proposed approach can account for the nonlinearity and ill-posedness inherent in this problem simultaneously. Nonlinearity is embodied in the relation between the scattered field and the target’s properties, and this relation is obtained through RVM training process. Besides, rather than utilizing regularization, the ill-posed nature of the inversion is accounted for naturally in respect that the RVM can produce probabilistic output. Once the training phase is completed, the proposed technique only needs computational time in an order of seconds to predict. Simulation results reveal that the proposed RVM-based approach can provide comparative performances in terms of accuracy, convergence, robustness, generalization and improved performance in terms of sparse property in comparison with the SVM-based approach.Lastly, various effects of presence of the wall, such as refraction, changing in speed and attenuation, can defocus target images, displace targets from their true positions and possibly produce false targets. In this dissertation, FDTD simulator is used to generate high fidelity through-wall radar data, and then the raw radar data are transformed into radar imaging using a back projection algorithm. It is effective to restrain the derivation of image modifying the time in imaging algorithm by the means of iterative method. It is shown that ultra-wideband through-wall radar can track and position targets inside a room with a wood wall.