This paper develops a data fusion algorithm of multisensor dynamic system based on filtering step by step.In distributed multisensor dynamic system when all of the observations aiming at the target are obtained

fistly we can predict the object state based on previous system information at this point and then use Kalman filtering and all of local observations to update the estimate value of object state in turn.Accordingly we can get a global fusion estimate value of object state based on the global information at that point.It presents the material form of this new algorithm and compares complexity of algorithm with traditional centralized data fusion algorithm.The computer simulation indicates that this algorithm possesses uniform estimate accuracy aiming at the object state with traditional centralized data fusion algorithm.