A Study Of Maneuvering Target Tracking Methods With The Quantized Measurement

As an important part of the target tracking and data fusion, maneuvering target tracking technology has been at the focus and cutting-edge research, and it is highly concerned by the domestic and foreign research institutions and scholars. With the rapid development of technology, and as a common non-traditional measurement in the communications and information processing, quantized measurement became one of the difficulties need to be solved in the current maneuvering target tracking. At the same time, how to apply the maneuvering target tracking technology to the quantized measurement will have a more long-term and wide-ranging significance. Maneuvering target tracking contains two main approaches. One is based on an adaptive model, and the other is based on a hybrid model. In this paper, the author mainly investigates these two maneuvering target tracking approaches with the quantized measurements.1. The causes and the current development of quantized measurements, and the development and application of maneuvering target tracking are studied. Then the non-traditional measurements, the error bounded interval model, the generalized likelihood of the quantized measurement, the model of the maneuvering target, and multi-model approach are introduced.2. A Singer model Kalman filter with the quantized measurement are studied and implemented. And a Singer model Gaussian-Mixture Probability Hypothesis Density (GMPHD) filter with the quantized measurements is presented. Taking a median of an interval and the approximate minimum mean square error (MMSE) method were used to convert quantized measurements into a point measurement approximately. These methods can apply to Singer model GMPHD filter. The algorithm can get a closed solution of linear Gaussian system. It can also handle measurements affected by three sources of uncertainty:stochastic, set-theoretic and data association. Simulation results verify the effectiveness of the filters. By comparison, the error with the quantized measurement is close to that of the point measurement, and the error taking the MMSE method is less than taking a median of an interval.3. The interacting multiple model particle filter (IMMPF) and the interacting multiple model box-particle filter (IMMBPF) algorithm with the quantized measurement are presented. This paper proposes an interacting multiple model particle PHD filter and an interacting multiple model box-particle PHD filter algorithm with the quantized measurement from the passive sensor. When a model index is set in the particle/box-particle state vector, the target state PHD is modeled with the particle/box-particle, and the particle/box-particle PHD filter are completed through the resampling and the input-interacting steps. The proposed algorithm can handle the nonlinear and non-Gauss multiple maneuvering target tracking. Under the same conditions, the target number estimated by the interacting multiple model box-particle PHD filter algorithm is almost closed to that of the interacting multiple model particle PHD filter algorithm. The number of box-particles used in the former is far less than the number of particles used in the latter.