Acceleration variance adaptive adjustment method of maneuvering target current statistical model has low tracking precision for weak maneuvering target and non maneuvering target. The paper puts forward a modified method. Discrete state equation of maneuvering target current statistical model is founded. Discrete observation equation of radar seeker is also founded. This paper puts forward an improved acceleration variance adaptive adjustment algorithm of unscented kalman filtering(UKF),which uses disturbance between radar seeker observational information and prediction value of position to self-adaptive adjustment acceleration variance.The simulation shows that modified CS_UKF algorithm has better tracking precision for weak maneuvering target, non maneuvering target and high maneuvering target.