Aimed at the problem that the traditional twostage Kalman filtering can only deal with single or part of complex noises, a twostage Kalman filtering fusion algorithm based on sequential decorrelation of noise is proposed in full consideration of the modeling of colored noise and four kinds of noise correlation in multiradar tracking system. First, this paper is to give the order of noise decorrelation which can effectively avoid the coupling of noise correlation, and then to obtain a target tracking fusion model uncorrelated among the colored noise, the process noise and the measurement noise by using the equivalent transformation technology. Finally, the centralized Kalman filtering fusion of multiradar measurement noise correlation system is realized by using the square root decomposition and the unit lower triangular matrix inversion technology.