The Algorithm Research Of Maneuvering Target Tracking Based On Multiple Models

Maneuvering target tracking is an important research direction in the field of target tracking, and it has a wide range of applications in the national defense scientific research and civil fields. Target tracking, it is an uncertainty typical problems. And these uncertainties mainly come from the maneuvering target model and the selection of filtering technology. Due to the unknown target motion state, the observation source is uncertainty, multiple objectives and environmental noise also influence target tracking. Finally, based on the multiple models algorithm, the problem of maneuvering target tracking was studied deeply.The main work of this paper and achievements are as follows: 1. This paper first introduces the basic principles of maneuvering target tracking, then introduces the target model and filtering algorithm for maneuvering target tracking. Establish an appropriate target tracking motion model is a prerequisite for accurate tracking. A variety of single models are respectively introduced, such as the constant velocity model, the constant acceleration model, turn model, current statistical model. Because a single model is not a good coverage of maneuvering targets, and therefore multi-model algorithm is studied. Another important aspect is to achieve maneuvering target tracking filter algorithm, the early use of a linear kalman filter, but mostly because the actual target is nonlinear state, and therefore the introduction of the extended kalman, unscented kalman, cubature kalman etc. nonlinear filtering techniques. Finally, using the same filtering techniques and different motion model algorithms, they were simulated and compared; and the same motion model, different filtering algorithm for the simulation and comparison, finally, through simulation experiments,this paper demonstrate the effectiveness of each method.2. This paper presents a new approach, which is the square root cubature kalman filter based on the recombined interacting multiple model, namely RIMM-SRCKF-FL. The realization process, based on the IMM algorithm, adopt the recombined weighted idea.In addition, this paper use the SRCKF in the filtering prediction method, which uses the spherical integral and the radial integral criterion. Compared to nonlinear filtering algorithm used in a wide range of UKF, it optimizes the UKF in point sampling strategies and weight distribution. Meanwhile, SRCKF introduced QR decomposition, avoided prescribing matrix operations, improve the stability of the filter. Finally,introducing the smoothing algorithm, the tracking accuracy can be further improved. The simulation results show that the proposed method can improve the real-time and accuracy of target tracking.3. Analysis of the defect interacting multiple model algorithm, and then study the variable structure multiple model algorithm. Focusing on the two kinds of variable structure multiple model, namely the adaptive grid models and the expected model augmentation. In AGIMM the adaptive grid models combined with SRUKF and SRCKF respectively, AGIMM-SRUKF algorithm and AGIMM-SRCKF algorithm are implemented, the filtering effect has been further improved. In the EMA, through the improvement of the acceleration model set, which makes the maneuvering target tracking accuracy have improved. For the quantitative methods of acceleration model set, presents a random intervals, the paper respectively adopt uniform quantization interval and random intervals, they are simulated and analysised. For the above research, this paper has passed the corresponding simulation and they have been verified. Finally, the thesis is summarized, and some further research issues are presented.