Research On Track-before-detect Algorithm For Radar Dim Targets Based On Particle Filter

Detection and tracking of dim targets is one of the key issues that modern radar have to face.When the energy of target echo signals is low or the background interference is very strong,the signal-to-noise ratio(SNR)of the target echo received by the radar sensor can be very low.Under such condition,the conventional track-after-detect(TAD)method based on single frame threshold decision is hard to undertake the task of detection and tracking.A new method called track-before-detect(TBD)provides an effective solution to this problem.TBD integrates detection and tracking,does not set the threshold for single-frame echo data,and carries out energy accumulation along the possible path of the target,effectively detecting and tracking dim targets with very low SNR.Among the many algorithms of TBD,the particle filter(PF)based TBD algorithm(PF-TBD)derived from Bayes recursive estimation theory is superior in performance,which is the focus of this paper.Firstly,this paper studies the theory of the standard PF algorithm as well as its improved version of free resampling –the Gaussian particle filter(GPF)algorithm.For the problem that Quasi-Monte Carlo(QMC)method used in the GPF algorithm improves the performance but increases the computational complexity,a simplified algorithm of linear transformation of the set of elementary particles in the QMC sampling process is proposed which is called SQMC-GPF for short.The simulation results show that the improved algorithm has comparable precision than the QMC-GPF algorithm and exactly a higher speed.Secondly,this article presents a new algorithm called RGPF-TBD where the GPF is combined with the RPF-TBD algorithm to solve the problems of low diversity and parallelism of particles caused by the resampling step of standard PF algorithm.And the detailed derivation is also given.The new algorithm inherits the advantages of the GPF algorithm,eliminates the need for a resampling step and has a higher parallelism,and the diversity of the particles is guaranteed,resulting in better detection and tracking performance.Then,QMC-RGPF-TBD algorithm is proposed by replacing the MC method in the RGPF-TBD algorithm with the QMC method and replacing the pseudo-random sequence in the RGPF-TBD algorithm with the super-homogeneous sequence(also called quasi-randomsequence)during the estimation process of birth density.This algorithm can improve the diversity and utilization of particles and lead to better detection and tracking performance.Simulation results show that the algorithm has a good performance in detecting and tracking compared with RGP-TBD and RGPF-TBD algorithms.However,it has higher computational complexity because of the introduction of QMC method.To solve this problem,the a new algorithm called SQMC-RGPF-TBD is proposed to simplify the QMC sampling in continuous dnesity estimation by linear transformation,and replace the various quasi-random samplings with only one quasi-random sampling in the estimation of the birth density.Simulation results show that the algorithm has good performance of detection and tracking and a faster speed.Finally,the target motion and radar measurement system are modeled,under which the algorithm of RPF-TBD,RGPF-TBD,QMC-RGPF-TBD and SQMC-RGPF-TBD are applied to the detection and tracking of dim targets,and according to the results,the four algorithms are compared and analyzed.It is concluded that the QMC-RGPF-TBD algorithm has the best detection and tracking performance when it does not consider the amount of computation.When the real-time performance and performance are required at the same time,the SQMC-RGPF-TBD algorithm is the first choice.