In order to improve low filtering precision and divergence caused by sensor faults in target tracking, an adaptive unscented Kalman filter(UKF) is proposed. In the filtering process, by applying an adaptive matrix gene for the UKF according to the adaptive estimation principle, the algorithm can adjust the covariance matrixes of the state vector and innovation vector in real time, which meets the optimal conditions of the UKF algorithm. Then, the filtering divergence is judged and restrained by taking some measures. Compared with the traditional and existing adaptive UKF algorithm, the filter accuracy and numerical stability are remarkably improved in this adaptive UKF filter algorithm, and an adaptive capability to deal with sensor faults is performed. Simulation results show the effectiveness of the proposed algorithm.