Research On Distributed MIMO Radar Track Before Detect Technology

With the development of stealth technology,more and more weapons and equipment have the stealth properties.Correspondingly the anti stealth systems and technology of the military power researches in full swing.The distributed MIMO radar system can use the space diversity to resist the target RCS glint and has the unique advantage in the anti stealth.In addition,as a multi frame data processing technology,which can effectively improve the SNR and improve the detection performance of weak targets,the track before detect technology has attracted more and more attention.In order to improve the target detection performance of distributed MIMO radar system,based on a thorough study of related key technologies,the paper focus on and discusses track before detect technology under different noise environment of Gauss,non-Gauss and other unknown statistical properties.The main work includes:A track before detection algorithm based on dynamic programming and successive target cancellation is proposed aiming at the problem of target detection and tracking under the condition of Gauss noise.Firstly,the dynamic programming is improved,where the possible tracks are selected by the statistics.Then the observation data is accumulated along the track.After threshold decision,the single target can be detected and tracked.Followingly,multiple targets can be detected and tracked according to the idea of successive target cancellation.The detection probability and the complexity of the algorithm are deduced and the simulation results show that the complexity of the proposed algorithm is moderate and the detection performance is better than the traditional single frame detection algorithm.A track before detection algorithm based on cardinalized probability hypothesis density is proposed aiming at the problem of target detection and tracking under the condition of non-Gauss noise.Firstly,a group of particles with weights is used to approximate the probability hypothesis density of the target.The particle is updated by the observation data of each channel after particle prediction.Then,the cardinality distribution is derived when the number of targets is limited,which realizes the recurrence of the cardinalized probability hypothesis density under non-Gauss noise.Finally,the number of targets is estimated according to the cardinality distribution.The target states can be estimated by K means clustering analysis of the approximate probability hypothesis density.Simulations show that the number of target estimation and tracking error of the proposed algorithm are significantly lower than that of the probability hypothesis density.A track before detection algorithm based on cost-reference particle filter is proposed aiming at the problem of target detection and tracking under the condition of unknown noise statistics characteristics.Firstly,the cost risk is designed with the unkown noise statistics characteristics to replace the original weight of the particle.Then,the particles are obtained through importance sampling.Re-sample after prediction and updating to prevent the particles from decaying.The target states are estimated by approximating the target posterior probability density.At last,a detector based on cost reference is designed to judge whether the target exists or not.The relationship between the cost reference detector and the generalized likelihood ratio detector is analyzed.The simulation results show that the detection performance of the proposed algorithm is worse than that of the traditional particle filter when the noise statistics characteristics is known.However,the proposed algorithm performs better than the traditional particle filter under unknown noise statistics characteristics.