When the statistical characteristics of the system noise and measurement noise are indefinite,the estimated values of the two kinds of noise based on Adaptive Kalman Filter (AKF) are relevant,which does not conform to the reality,and may lower the precision of the filter.In order to solve the problem,Improved Adaptive Kalman Filter (IAKF) is proposed.The state space equation of a highly-integrated navigation system is built up.Principle of adaptive Kalman filter is introduced,and the reason for the relevance of two kinds of estimated values is given.Based on that,the new algorithm is used to estimate the system noise and measurement noise on-line,and the shortcoming of the original algorithm is overcome.Semi-physical simulation imaging experiment is designed to verify the new algorithm.The result shows that:The estimation accuracy of the improved algorithm with unknown system noise and measurement noise corresponds to that of the original algorithm with known system noise and measurement noise,which verifies the feasibility of the improved algorithm.