Underwater Multi-target Tracking Technology Of Active Sonar

In the process of target detection,active sonar inevitably produces measurement errors due to the influence of sensor noise,accompanied by a large number of false alarms and missed detection.In sonar signal processing,target tracking technology is used to estimate the state of multiple targets and identify targets.The traditional multi-target tracking algorithm represented by Multiple Hypothesis Tracking(MHT)has the disadvantages of data association combination explosion,which is not suitable for the real-time tracking of multi-targets with dense clutter.Therefore,in the background of active sonar underwater multi-target tracking,a real-time calculation method of Rao-Blackwellized Monte Carlo Data Association(RBMCDA)algorithm,and the Probability Hypothesis Density(PHD)are studied in this paper.This article first analyzes the characteristics of underwater multi-target tracking of active sonar.The state vector is usually composed of target coordinates and velocities.The observation is composed of bearing and distance,and the observation equation is nonlinear.Thus,the Unscented Kalman Filter(UKF)is used for single-target state filtering in nonlinear environment.Then the RBMCDA algorithm based on Rao-Blackwellized particle filter is studied.This algorithm uses the Sequential Monte Carlo(SMC)idea to give a recursive form of particle filtering for data association,avoiding the complex data association combination,by which the tracking acts well.However,the tracking results can only be given by the particle with the maximum weight at the end of the tracking,which cannot be given in real time.Thus,it is a batch processing algorithm.This paper proposes a new algorithm based on density clustering and RBMCDA,named as C-RBMCDA.It conducts density clustering on the target states under each particle at each moment,takes the cluster centers as the estimation of targets' state,and output the target trajectory in real time through the target label management.In the simulation experiments,UKF is embedded into the new algorithm.The experiment results show that the new algorithm performs well in real-time multi-target tracking,and it has better robustness than RBMCDA in the case of prior information mismatch.Multi-target Bayesian filter based on Random Finite Set(RFS)brings a new solution to multi-target tracking,skillfully avoids the data association in the traditional multi-target tracking.As the first-order statistical moment approximation,the PHD filter has two engineering methods: SMC approximation and Gaussian Mixture(GM)approximation,which has been widely used.In order to eliminate the reliance on the priori information of the new targets,this paper proposes that the priori position of the new target is calculated from the observation data at the last time.Since the PHD filter is just a multi-target state estimator,it cannot give target identification.In this paper,label information is added to each Gaussian component in GM-PHD filter,and the target state with an identity is output through label management.Simulation results show that the algorithm is effective in high clutter ratio,multitarget trajectory crossing and target maneuvering environment,which is more efficient than RBMCDA algorithm.