Research Methods Of Passive Sonar Detection And Multi-sonar Tracking

Nowadays, due to a variety of military and civilian needs, detecting and trackingunderwater remote passive target becomes the research focus. For passive system, due toincomplete observation, single sensor can’t fulfill the tracking task. Merge the azimuthinformation of multi-sensor can guarantee the tracking task accomplished in the case ofobservation station remain stationary. The research topic in this paper is detecting andtracking passive moving target in the case of multi-node sensor.The research can be divided into two main parts, find the presence of target and estimateits parameters is the main content of the first part, tracking the target with the obtainedparameters is the main content of the second part.In previous section, firstly, two classical beamforming algorithms are researched, theyare conventional beamforming and MVDR, in order to reflect the respective strengths ofthese two algorithms, simple numerical simulations are used to compare these two algorithms.For wideband signal, the delay and sum implement form of conventional beamforming can beused, but MVDR unsuitable for wideband signal, so in this paper, broadband incoherentMVDR algorithm is researched to meet the requirements of wideband signal, conventionaldelay and sum beamforming algorithm and broadband incoherent MVDR algorithm aresimulated in the case of passive single target and passive multi-objective. Secondly, theMVDR algorithm has poor robustness, can easily affected by the error of steering vector andcovariance matrix, so in this paper, two improved methods are researched to increase therobustness of MVDR. Finally, aiming at the disadvantage of conventional beamforming thatits azimuth distinguish ability is restrainted by Rayleigh limit, delay and sum virtual arraytechnology is researched in this paper to improve the azimuth distinguish ability. But aproblem is found in processing the experiment data that this method needs high samplefrequency, in practice, the sample frequency can’t meet the requirement, inspired by theapproach of wideband signal, broadband incoherent virtual array technology is researched inthis paper, this method can achieve accurate azimuth estimation in the case of lower samplefrequency.In the latter part, firstly, a series of target tracking basic theory are introduced, includingthe moving model of target, the observation model of sensor and accuracy evaluation index.Secondly, a class of target tracking methods that based on kalman filter are studied. Mainlyhave three algorithms, they are least squares kalman filter, extended kalman filter and interacting multiple model extended kalman filter, and these three methods are simulated bycomputer. Finally, a class of target tracking methods that based on particle filter are studied.Mainly have three algorithms, they are standard particle filter, improved particle filter basedon uniform resample and interacting multiple model particle filter, also these three methodsare simulated by computer.At last, aiming at particle filter algorithm, several problems which personally think to beresolved are proposed.