Multi-aspect SAR Imaging And Feature Extraction

Multi-aspect SAR system is composed of several SAR sensors which interrogate the same scene or object from different aspect angles. The multi-aspect SAR data contains rich information for radar imaging and feature extraction because of angular, frequency and waveform diversity. Moreover, comparing with MIMO radar or radar sensor network(RSN), multi-aspect SAR data is obtained easily from existing SAR systems. So, it is valuable to investigate multi-aspect SAR imaging and feature extraction.This dissertation exploits target information, which is embeded in multi-aspect SAR data, by developing imaging and feature extraction algorithms. Some new imaging and feature extraction methods are proposed for the purpose of resolution enhancement, accuracy improvement and robustness upgrade. In addition, the space-variant characteristic of target scattering is another topic of this paper. Considering the coherent of radar measurements, the imaging and feature extraction algorithms are applied at the signal-level.Because of the discontinuity of multi-aspect SAR sampling, the classic SAR imaging algorithms may rusult in high sidelobe and artifact. In chapter 2, a 2-D matched filtering method is proposed for multi-aspect SAR imaging. This method forms 2-D matched function by adjusting the reference point, it solves the focusing and discontinuity sampling problem, it is also applicable for other non-uniformly sampling data. This method offers great potential for better resolution over classical imaging algorithms, it can also synthesize target feature from different sensor-target orientation. In order to improve the efficiency of this algorithm, the ROIs are found firstly by single-aspect SAR imaging, then these ROIs can be imaged by multi-aspect SAR imaging algorithm.The radar imaging algorithms are classified as matching filtering and parameter estimation method. In chapter 3, the Gapped-Data Amplitude and Phase Estimation(GAPES) are modified for multi-aspect SAR imaging. This algorithm don’t needs the parametric model of object. In addition, there is no necessity find the missed data and avoid the error induced by the interpolation. The new approach is not only able to improve the estimation accuracy of location and amplitude, but also improve the resolution. Numerical experiments are provided to demonstrate the performance of the algorithm and to show the advantages of multi-aspect SAR data to reconstruct object.In chapter 4, the multi-aspect SAR imaging based on compressive sensing is studied. From the view of information theory, the sampling rate of multi-aspect SAR do not meet the Nyquist rate. The recently introduced theory of compressed sensing(CS) states that it is possible to recover such sparse or compressible signals accurately from a highly incomplete number of samples with high probability by solving a convex l1 optimization problem, even with much fewer measurements than what is considered to be necessary according to the Nyquist sampling theorem. Motivated by CS theory, a few efficient schemes are proposed for multi-aspect SAR imaging. First, A new criterion, the point ambiguous function(PAF), is proposed to analyze the performance of compressed synthetic aperture radar(SAR) imaging. This criterion is deduced from the mutual coherence of compressed sensing(CS) theory, which characterizes the ambiguity of adjacent scatterers. With this criterion, the present study analyzes the factors that dominate the performance of compressed SAR imaging. For compressed SAR, PAF is dominated by the transmitting signal and the data collection geometry. Specifically, the robustness of compressed SAR can be improved by transmitting stochastic waveform and sampling with random spatial position. Second, the wide-angle SAR imaging is studied. The model mismatch function(MMF) is presented to study the instability which is resulted in by the mismatch of point scattering model assumation for wide aperture SAR. By these analysis, the generalized likehood ratio test(GLRT) is appled to relieve the instability of imaging which is caused by model mismatch. Finally, with this PAF criterion, the present study analyzes the factors that dominate the performance of CS-based Multi-aspect SAR imaging. The angular-frequency-diversity can improve the robustness and resolution of 3D imaging. The PAF provides a quantitative tool to study the CS-based radar imaging, and can be utilized to design CS-based radar imaging system.In order to improve the precision and robustness of parameters estimation, the multi-aspect SAR data is applied to estimate model parameters of attributed scattering center model. Chapter 5 considers the parameters estimation as a sparse signal reconstruction problem, and propose parameters-sequential algorithm to relieve the computational complexity. Two factors are studied. First, the aspect and frequency diversity can improve the performance of dictionary matrix. Second, in order to reduce the algorithm complexity, the parameters estimation procedure is realized sequentially. The initial imagery is reconstructed by dictionary matrix which is built up by the ideal scattering point model. The model order, type and location of scattering center are established primarily by energy segment of initial imagery. Then all parameters are estimated over again based on the dictionary matrix which is built up by the prior estimation. The feasibility and robustness of algorithm is validated by numeric simulation.