Research On Target Detection In Passive Radar Based On Sparse Representation

In contrast to traditional active radars,the passive radar has advantages of covertness,availability of abundant illuminators of opportunity(IOs),low cost,and easy deployment,which has been extensively employed for practical civilian and military applications.Over the past decade,digital information technology has explosively developed.Compared to traditional analog radio signals,the ambiguity functions of digital broadcasting,television,as well as communication signals are generally of the “thumbtack”shape with better sparsity.This dissertation studies the target detection based on sparse representation for diverse passive radar systems and proposes several novel detection algorithms without requiring signal reconstruction,which effectively reduces the computation complexity and circumvents the problem of high sampling rate,excessive data volume,and insufficient storage space required in real-time radar signal processing.The main work and contributions of this dissertation are summarized as follows:1.For bistatic passive radar systems,the multi-target detection with direct-path interference and multi-path clutter in target echoes is studied.Based on the interference characteristics,we design a random compressive measurement matrix,which is capable of removing the interference without sacrificing compressive observations.Furthermore,an orthogonal matching pursuit(OMP)detection algorithm is proposed built on the residual distribution and analyzed by deriving the closed-form expression of the false alarm probability.Moreover,considering the residual interference,a compressive constant false alarm rate(CFAR)detector is proposed by designing a determined compressive matrix exploiting the sparse property of the target echoes.Simulation results demonstrate that the proposed algorithms outperform benchmark algorithms.2.For the scenario of a multistatic passive radar with single IO and multiple receivers,the compressive detection problem is studied considering the direct-path and multi-path clutter interference in the target echoes.For a target that is determined to exist but unknown in the detection units,the relative time delays and Doppler shifts of diverse target echoes are first calculated by employing the spatial position information of the IO and receivers,following which the compressed measurement matrices associated with all receivers are jointly designed.Furthermore,a pair of generalized likelihood ratio test(GLRT)based compressive subspace detectors are proposed in the case of known/unknown noise variance,respectively.Additionally,closed-form expressions of false alarm probability of the proposed detectors are derived.Finally,numerical simulations are provided to verify the detection performance of the proposed algorithms.3.For the scene with a multistatic passive radar with single IO and multiple receivers,the target detection without interference is studied.Firstly,measurement matrices are jointly designed by exploiting the correlation of the target echoes.Secondly,theoretical analysis proves that the compressive observations under the binary hypothesis both follow complex Gaussian distribution with identical mean but diverse variances.Based on this property,a pair of multiple-measurement-vector(MMV)based detection algorithms are proposed in cases of known and unknown noise variance,respectively.Compared to the compressive subspace detector,the MMV algorithm has a lower complexity.Simulation results show that the MMV detector outperforms the compressive subspace detector in a moderate range of signal-to-noise ratio.4.For the case of a multistatic passive radar system with multiple IOs and a single receiver,compressive target detection in the single frequency network(SFN)is studied.Two situations with known/unknown IOs' locations are considered.Specifically,when IOs' positions are prior,a pair of SFN-based compressive subspace detectors are proposed under known/unknown noise variance.By contrast,when considering unknown IOs' locations with an unknown target support set,an order-statistic based orthogonal matching pursuit(OSOMP)detection algorithm is proposed.Additionally,the analytical expression of the false alarm probability of the OSOMP algorithm is derived.To reduce the complexity,we further discuss the minimum number of iterations required by the OSOMP algorithm to achieve the desired probability of detection and false alarm.