Research On Multi-frame Detection Techniques For Over-the-horizon Radar Targets

The track before detect(TBD)technology is different from the traditional radar target detection and tracking process.By accumulating the echo data of multiple coherent accumulation periods,higher tracking accuracy can be achieved,and at the same time,low signal to noise ratio can be achieved.Weak targets to detect.Among the numerous TBD implementations,the particle filter-based TBD(PF-TBD)technique can be applied when the time-varying system is non-linear and non-Gaussian,and has a wider application range.For over-the-horizon radar(OTHR)observations,the characteristics of weak targets are often presented due to interference,noise,and clutter.Therefore,it is very meaningful to study the track before detect technology based on particle filtering.The specific research content of this paper is as follows:The principle and implementation process of particle filter algorithm based on Bayesian estimation theory are studied.The two classic PF-TBD algorithms standard PF-TBD(SPF-TBD)and efficient PF-TBD(EPF-TBD)are compared and analyzed under different SNR conditions.For the track before detect problem of maneuvering targets,based on the multi-model PF-TBD algorithm,the motion of the maneuvering target is summarized into several typical motion models.In addition,a new motion model variable is added to the target state vector to realize maneuverability target detection and tracking.An echo model matching the actual OTHR system is established.In order to solve the particle degradation problem that restricts the performance of PF-TBD algorithm,particle swarm optimization(PSO)step is added to the particle sampling process of EPF-TBD.By using the random part of PSO algorithm,particles can be maintained diversity.In addition,by increasing PSO steps,particles can move to regions with large likelihood function.Under the OTHR echo model,the corresponding target state model and observation model were set up,and the performance of EPF-TBD and PSO-EPFTBD algorithms was compared.Under the OTHR signal echo model,the PF-TBD algorithm under two kinds of clutter environments was studied.Firstly,using the zero memory non-linearity(ZMNL)to generate the Weibull clutter sequence,the modified likelihood function and particle weight are derived in the Weibull clutter environment.The comparison with the original SPF-TBD proves the feasibility and effectiveness of the new algorithm.The use of spherically invariant random processes method(SIRP)to generate K-distribution clutter sequences,and a method to realize PF-TBD in the background of K-distributed clutter is studied.The simulation results show that the algorithm performance has been partially improved.