Aspect Dependent Efficient Multipath Ghost Suppression In Through The Wall Radar With Compressive Sensing
Ali H. Muqaibel, عبدالطلب عبدالله
كلية الهندسة-جامعة الملك فهد للبترول والمعادن · السعودية
Recently, Through-the-Wall Radar Imaging (TWRI) society has witnessed a dramatic milestone after successful application of Compressive Sensing (CS) and sparse image reconstructions. However, multipath propagation is still a challenging problem. Multipath creates ambiguities in the measurements which result in ghost targets that can be confused with genuine targets. In this dissertation, multipath ghost suppression method based on Aspect Dependent (AD) feature is developed under CS framework. A measurement procedure is proposed using strategically selected duo-subaperture to realize the AD feature. The corresponding subimages are fused using either weighted sum or harmonic mean based techniques proposed in this work. Apart from using small fraction of the measurement, the knowledge of the reflecting geometry is not needed making the proposed method feasible. The results based on MATLAB and experimental data returned Target Signal-to-Clutter Ratio (TSCR) of 97.3dB and Target Relative Clutter Peak (TRCP) of 23.7dB which are relatively higher compared to the existing methods.It was found that under given conditions, the optimal configuration to suppress multipath ghost is to divide the aperture into two, reconstruct the respective subimages and then combine them via effective fusion technique. Our contribution answer the questions as how many subapertures and how to divide the aperture to effectively suppress the ghosts.To enhance the performance of the proposed method, sparse subarrays based on Pythagorean triple are proposed and evaluated. Two imaging modalities are examined: subarrays imaging followed by image fusion and imaging using combined subarrays. In each case, three configurations are discussed: Pythagorean-based Interlaced Sub-Apertures, Pythagorean-based Displaced Sub-Apertures (PDSA) and Spatial Orthogonal Coprime Arrays. The effectiveness of the subarrays is examined using MATLAB simulation and specialized EM propagation software. Imaging using duo-subarrays with PDSA configuration gives a TRCP of around 5dB higher than random selection.In some applications, the conventional signal model does not hold because the target occupies more than one pixel. This work, extends the existing model to incorporate extended targets. The ground truth is modeled to contain complex shaped targets forming blocks of pixels of unequal sizes with unknown reflectivity distributions and the recently proposed blockSABMP algorithm is implemented in TWRI problem. The approach is scrutinized under different scenarios such as complex shaped target, non-homogeneous scene and Rayleigh distributed target. For all scenarios, it returns the best Earth Mover’s Distance values and Normalized Mean Squared Errors compared to the existing algorithms.Moreover, this work proposes a single antenna localization scheme for indoor target utilizing virtual radars. The fact that multipath returns are AD, the radar is located such that it receives minimal returns. This reduces wall ambiguity significantly and simplifies the process of associating the multipath to their respective walls. Simulation results show that the proposed method is robust for a good range of timing errors.