Research On Filter Design And Implementation Methors For Moving Targets Tracking

With the further research of complex systems,and higher requirements levels about estimation and control,target tracking has been more focused by experts and specialists of concerned field.On the one hand,in practical engineering application,systematic nonlinear characteristic was inevitably prominent through coordinate system transformation based on target motion modeling and measurement modeling,besides transformation and registration of space with distributed measurement information.What's more,owing to traditional nonlinear methods are generally based on linearization and Gaussian assumption,the precision and application of nonlinear filter are affected in a certain extent.On the other hand,in accordance to the moving target tracking in complex environment,multi-target tracking is research priority under clutter environment and uncertain movement patterns in the field of target tracking at home and abroad.Therefore,further improvement of the precision of target state and target number estimation,as well as the reasonable selection and optimization of filter model,became the key problems desiderated to be solved.In this paper,the related research work is carried out on the basis of the basic filtering methods under the system of multi-sensor fusion,multi-target tracking and multi-model transformation,and the main contributions are as follows:Aiming at the consistency deviation occurred in virtual measurement sampling process on account of measurement noise uncertainty,a novel multi-sensor ensemble Kalman filtering algorithm based on Metropolis-Hastings sampling is proposed.Firstly,combined with the physical properties of multi-sensor measurement system and the generation mechanism of bootstrapping measurement in ensemble Kalman filter,multi-sensor bootstrapping measurement set is structured.Secondly,through solving the likelihood of multi-sensor bootstrapping measurement and designing the probability function of measurement acceptance,validation measurement from multi-sensor bootstrapping measurement set is confirmed by Metropolis-Hastings sampling strategy.The new method corrects the consistency deviation appearing at bootstrapping measurement by means of the extraction and utilization for the redundancy and complementary information in multi-sensor measurement,and improves the filtering precision for the estimated system state.Under clutter environment,multi-target tracking system modeling is nonlinear,and measurement noise is random.All these negatively affect the precision of state estimation and target recognition.Thereby,I bring forward a filtering algorithm,called central difference Gaussian mixture probability hypothesis density filter based on consistent fusion and measurement bootstrap.First of all,to process the nonlinearity of system,I proposed to design a filter,which can dynamically integrate central difference Kalman filter and Gaussian mixture probability hypothesis density filter;Secondly,through the extraction and utilization of redundant and complementary information in multi-sensor measurement,we can effectively improve the stability and reliability of the measurement.Besides,construct virtual multi-sensor measurement based on bootstrap sampling,and then referring to the similarity of measured data,we can construct consistency distance and consistency matrix.Finally,via weighted fusion,we can realize the reasonable utilization of the virtual multi-sensor measurement information.The selection and optimization of model filter directly affect the precision of motion pattern identification and state estimation in maneuvering target tracking problem.Aiming at the performance improvement of model filter,a novel maneuvering target tracking algorithm based on central difference Kalman filter in observation bootstrapping strategy is proposed in this paper.The framework of interactive multiple model(IMM)is used to realize the identification of motion pattern,and the central difference Kalman filter(CDKF)is selected as the model filter of IMM.Considering the advantage of multi-sensor fusion method in improving the stability and reliability of observation information,and the hardware cost of observation system if multiple sensors are adopted,meanwhile,learning from the data assimilation technique in ensemble Kalman filter(EnKF),the bootstrapping observation set is constructed by integrating the latest observation and the prior information of observation noise.On that basis,these bootstrapping observations are reasonably used to optimize the filtering performance of CDKF by means of weight fusion way.The object of new algorithm is,without increasing the number of physical sensors,to improve the tracking precision of observed target by multi-sensor fusion method.The theoretical analysis and experimental results show the feasibility and efficiency of proposed algorithm.