Study On Sparse Signal Processing For Radar Detection And Imaging Application

The theory of sparse signal processing, especially the establishment and rapiddevelopment of compressive sensing (CS) in recent years, has provided a new direction forthe application of sparse signal processing technology in the solution of practical radar signalprocessing problems. Sparse signal processing technology provides us with a framework toaccurately obtain the target information of the radar with a few echo data. According to theresearch at home and abroad, sparse signal sampling and recovery technique own a greatpotential in radar application. The radar system based on CS can simplify the design of radarhardware, improve the resolution, shorten the time for obtaining data, reduce data memorycapacity and reconstruct partial signals. This dissertation enumerates some applications of CStheory in radar target detection and imaging, makes a research on some new methods of radarsparse signal processing by means of conceptual analysis, theoretical derivation, real dataverification, etc., and finally studies the design of the radar measurement matrix based on CSand the real-time processing circuit design of its optimization algorithm. The main content ofthis dissertation is summarized as follows.The dissertation, on the basis of the practical applications of radar sparse signalprocessing and combining the basic theory of CS, analyzes sparse expression of radar echosignals, measurement matrix design, optimization problem solution and clutter processing.Over-the-horizon radar(OTHR) operating in the high frequency (HF) band can providelong-range detection of targets in large surveillance areas. However, OTHR signal is usuallycontaminated by transient interference, such as lightning, meteor trail echoes and man-madeimpulse interference. To filter out these interferences is available, while it destroys theintegrity of signal in time domain. What’s more, modern OTHR system is usually required tobe multi-mode and able to detect multi-targets simultaneously under different backgrounds.To satisfy these requirements, incontinuous sampling mode is usually applied resulting inparts of signal missing. In this paper, a novel algorithm is proposed to reconstruct theintegrate spectrum of OTHR with partial signal. The spectrum reconstruction problem isshifted into a problem of a norm1constrained optimal problem. By solving this problem withgreedy algorithm efficiently, the integrate frequency spectrum of OTHR signal isreconstructed optimally.The dissertation studies the range-Doppler two-dimensional (2D) high resolutionconstruction of HF radar randomly hopped frequency signals. In a congest frequency band,it’s hard for a HF radar to find continuous broadband to realize high range resolution. Thus, the dissertation proposes to use randomly-distributed sparse frequencies to realizerange-Doppler2D high resolution processing. Because of the absence of some frequencies,high sidelobes and grating lobes will arise in the target’s range profile. To solve this problem,a range-velocity redundant time-frequency dictionary matrix is built to realize thetwo-dimensional high resolution target construction through optimization solution. For theproblem of the great coherence of the dictionary matrix of the moving target, the dissertationalso suggests to reduce the size of the dictionary by speed estimation, so as to improve itsorthogonality of the dictionary’s items. This method is able to enhance the recovery accuracyand stability of CS algorithm, and meanwhile improve the calculation efficiency ofoptimization solution.A novel velocity ambiguity resolving method is proposed for moving target indication(MTI) radar, in which CS is applied to recover the unambiguous Doppler spectrum of targetsfrom the randomly pulse repetition frequency (PRF)-jittering pulses. Aiming at the issue thatstrong clutter affects the recovery accuracy of signals, the Doppler ambiguity resolvingmethod is improved. Firstly a group of reference pulse is transmitted to obtain clutterspectrum estimation. Then weighting in CS recovery is utilized to suppress the clutter byusing the estimated clutter spectrum.In this dissertation, we present an algorithm for inversed synthetic aperture radar (ISAR)imaging with super resolution by combining CS and bandwidth extrapolation (BWE)technique. For ISAR imaging, the backscattering field of target is usually contributed by afew strong scattering centers, whose number is much less than that of image pixels. Thus, CSis intuitively suitable for constructing super resolution ISAR image. According to CS theory,the number of extracted dominating scatters relies on the signal length, which indicates that ifonly limited data is available, it is difficult to generate dense ISAR image robustly by CS, andsome signal components tend to lose. To soften this constraint, BWE is combined with CSimaging to increase the degree of signal freedom while preserving its coherence. A refinedCS-based formation for ISAR image-resolution enhancement is then developed. Both real andsimulated data experiments are performed to evaluate the proposed approach, and an exampleof using this technique demonstrates the enhanced image resolution in application ofmaneuvering target imaging.The dissertation also studies the design method of radar measurement matrix circuitbased on CS and the circuit realizing method of Orthogonal Matching Pursuit (OMP)optimization algorithm. The design of the random measurement matrix and the highlyefficient solution of the optimization problem are crucial for the successful application of CStheory in real-time radar signal processing. The design of the random measurement matrix involves the change and control of radar parameters, in this dissertation, the design method ofusing Field Programmable Gate Array (FPGA) and digital to analog converter to realize radarrandom measurement and the method of using FPGA and parallel-to-serial converter torealize the random modulation of broadband sparse signals are proposed. To solve theoptimization problem of sparse signal processing, a high-paralleled and deep-pipelinedcalculation circuit is designed on the basis of the high-performance FPGA platform developedby the lab. It compares the efficiency of the least squares problem based on ModifiedGram-Schmidt(MGS) QR decomposition and Conjugate Gradient(CG) iterative method. Thecircuit designed in the dissertation can be directly applied in real-time radar sparse signalprocessing.