Multi-sensor Information Fusion Methods Based On Two-stage Kalman Filter

With the development of science and technology, the monitoring technology by single sensor gradually cannot meet people’s requirements. Multi-sensors technology is playing an increasingly important role in fields of the social services, the industrial and agricultural,especially the national defense and military. One of the typical multi-sensors systems is the multi-radar networking system, in which, there are relatively fixed system errors, due to the discreteness, ageing, drifting of elements and the various environmental factors, and the fusion performances do deteriorate to some extent. Therefore, the estimation and compensation of system errors should be dealt with before tracking targets by the multi-radar networking system, which is named registration. A common registration method is to jointly estimate the system state and the system error, and a effect method to solve this problem is the two-stage Kalman filter. And the information fusion problems are addressed in this paper, and the main works can be summarized as follows:Firstly, in the parallel centralized fusion framework, an augment measurement is expanded by each sensors’ measurements, and the corresponding measurement function is also given. On this basis, the fusion estimates of the system state and the system error are solved by the single-sensor two-stage Kalman filter. The simulation results prove the effectiveness and the feasibility of the given parallel centralized two-stage Kalman fusion filtering method.Secondly, in the parallel distributed fusion framework, a parallel distributed two-stage Kalman fusion filtering method is presented in this paper, in which, the fusion parameters of the local estimates of all sensors are given by using the local estimation error conversances.Then, the global fusion estimate can be obtained. The simulation verifies the fusion filtering superiority of the presented parallel distributed two-stage Kalman fusion filtering method.In the parallel fusion frameworks, all measurements or local estimates are required to be obtained by the fusion center before the fusion process. The sequential fusion filteringmethods can deal with the measurement or local estimate once it is received by the fusion center. And the real time performance can be improved than the parallel fusion filtering methods. Therefore, the sequential centralized two-stage Kalman fusion filtering method and the sequential distributed two-stage Kalman fusion filtering method are also proposed.Thirdly, a sequential centralized two-stage Kalman fusion filtering method is given by filtering the measurements sequentially, in which, the estimate of current system state can be updated. The global estimate of system state can be obtained by dealing with the last measurements. The sequential centralized two-stage Kalman fusion filtering methods are with real-time performeance and less calculate requirement.Fourthly, a sequential distributed two-stage Kalman fusion filtering method is given by fusing the local estimates sequentially. The global fusion estimate can be obtained by fusing the last local estimate. The real-time fusion performance is shown in the numerical simulation section.