Research On Underwater Weak Target Tracking Algorithm

Passive sonar detection and tracking is widely used in production,military and other fields because of its excellent concealment.With the improvement of underwater target stealth technology,passive tracking systems often face harsh scenes with low signal-to-noise ratio.This thesis focuses on the problem of underwater weak target tracking,and studies the trackbefore-detect algorithm suitable for passive sonar under the framework of random finite set theory.First,considering the tracking problem of low signal-to-noise ratio targets,this thesis studies the conventional multi-Bernoulli filter for track-before-detect.Considering that the conventional algorithm is not suitable for passive sonar measurement,this thesis uses the weighted form of the MUSIC spatial spectrum as a pseudo-likelihood ratio function to studies the approximate multi-Bernoulli filter based on MUSIC.Aiming at the slow response speed of the algorithm to the target birth,the target birth model driven by measurement is studied.The simulation verifies that the research algorithm has better tracking performance under low signal-to-noise ratio than the traditional algorithm,and the improved target birth model can reduce the response time of the algorithm to the birth target.Afterwards,we study the filtering algorithm in the scene where the target azimuth trajectory crosses.Since the approximate multi-Bernoulli filter based on MUSIC cannot output the trajectory label,this thesis studies the track-before-detect algorithm based on the label multiBernoulli filter.Aiming at the problem that the original algorithm cannot solve the problem of trajectory crossing,the original algorithm is improved by combining the method of selective update.Simulation and actual data processing prove that the proposed algorithm can maintain the target trajectory well and has better track output and maintenance ability compared with the traditional algorithm in the scenario of target azimuth trajectory crossing.Finally,the track-before-detect smoothing algorithm based on label multi-Bernoulli filter is studied to further improve the accuracy of target state estimation.In this thesis,the forwardbackward smoothing algorithm based on label multi-Bernoulli filter is studied.Aiming at the problem of too long calculation time,the original smoothing algorithm is improved by the fast N-body Monte Carlo method,and the corresponding fast smoothing algorithm is proposed.The simulation proves that the smoothing algorithm studied further improves the tracking accuracy of the original tracking algorithm,and the fast smoothing algorithm studied can greatly reduce the calculation amount of the algorithm while ensuring the smoothing accuracy.