Abstract: The estimate accuracy of traditional Kalman filter depends on the state model of targets.When the state of targets would not be described by a precise mathematics model or a line state space model,the result will be volatilization.The paper provides a new algorithm by introducing GM(1,1) into Kalman filter.In course of Kalman filter,the forecasting value depends on the state model of targets no longer.The next value is forecasted by using a few forward estimated values with a grey differential equation.Not only its accuracy is higher,but also its performance is better.Especially,during target maneuvering,the result of the filter is better than that of the traditional method.Experiment results indicate that the way of maneuvering target tracking is feasible.