Research On Multisensor Observability And Spatial Registration

State estimation and fusion for target tracking is attractive in civilian and military field.With the sharp increase complexity of system application,the existing state estimation fusion theory is difficult to meet the growing demand of actual engineering.The traditional Kalman theory turns to get performance degradation or theory inapplicable when the filter model can't fit the real system well,especially the challenges of traditional evaluation criteria in real-time self performance evaluation.Focus on this,some researches have studied the performance evaluation of Kalman filtering theory by observable degree.Based on this work,some new improved Kalman filtering method have been designed and achieved well performance.Meanwhile,the problem of sensor registration in target tracking network has not yet been solved well.Therefore,the study of Multi-sensor observability analysis and space registration have great theoretical significance and applicable value.The main innovative of this paper includes as the following three aspects:(1)Study the observability analysis method based on Cramer Rao bound and least square method.Based on the existing research of observable analysis with least squares,to further consider the system contains process noise,in this paper by introducing the Cramer-Rao bound to analyze the existing method,revealing the relationship between them,and proposing an improved observability analysis method.(2)Study the observability degree of multi-sensor systems.Focus on Multisensor network,according to ODAEPM,study the observability degree of multisensor systems.This work is focus on the expression of observability degree between fusion center and the local sensors,also study the relationship between different fusion methods.(3)Propose a two step sensor registration method based on the unit quaternion.By transforming the transformation function,separating the different type of system bias and introducing the unit quaternion method.The new method has a higher estimation performance Compared with the existing methods,and also it can be applied to registration problem with large attitude bias.