Two-Dimensional Bearings-Only Target Tracking Algorithm Research And Application

Passive linear algorithm can not meet the high performance about bearings-only trackings which are nonlinear, non-Gaussian model.In such circumstances,particle filter is widely used in bearings-only target tracking. But the choice of the initial particle have an important impact to the particle filter result.In response to this situation,the paper discussed the sea surface situation that two-dimensional BOT-target motion analysis.First,we made an analysis and simulation about the classical methods which are recursive least squares(RLS) method and particle filtering(PF).And then,integrated the advantages of both, the initial values using recursive least squares estimation are estimated as the initial probability density of particle filter distribution. Finally, an improved method for the actual simulation particle filter tracked the BOT. This new method not only speed up the convergence rate of the particle filter, but also improve the stability of the particle filter. This method can overcome the particle filter method's insufficient that influenced by initial value.At the same time,the method have a practical significance about long-range BOT tracking.The main framework of the paper as follows:Chapter 1, General introduction the development of tracking the status quo. Described the paper's research and the main strategy.Chapter 2, Making two assumptions in order to facilitate further research topics, and introduces bearings-only target tracking theory, research methods and evaluation criteria. Chapters 3 and 4 introduced the least squares estimation method and the target trajectory initial particle filter recursive estimation method. Also introduced the principle and formula, and obtained experimental study the advantages and disadvantages of both approaches.Chapter 5, According to the previous experiment, an improved method to deal with BOT tracking problem is proposed. First, analysis the actual measured sea trial data, filter, get the data closed to the true azimuth measurement; then estimate the initial position and velocity; and use them as a PDF of the particle filter sampling. Finally, making the particle filter calculation and obtain convergent results.Chapter 6, Propose a new optimization method - auxiliary particle filter. The basic idea: before the particle filter sampling re-sampled first. The predicted likelihood large particles at t-1 time will be extended to t time used t time information, Thus increasing the diversity of particles, reducing the importance of the right variance. Through experimentation and simulation, we Discovered that auxiliary particle filter(APF) reduced the RMSE to a certain extent.,Chapter 7, The conclusions and contributions of this paper are summarized and highlighted. Moreover, some problems for further study are suggested as well.