Considering the filter divergence problem, a STUPF (Strong Tracting Unscented Particle Filter) algorithm which combines STF (Strong Tracting Filter), UKF (Unscented Kalman Filter) and UPF (Unscented Particle Filter)[1] is proposed. The STUPF algorithm improves the tracking performance of UPF filter by introducing the fading factor K to adjust the filter gain to reduce the weight of the old data, improve the weight of the new data. The traditional re-sampling algorithm can solve the problem of degradation, it is easy to cause the particle depletion; Despite less particle depletion, the extended particle filter algorithm EKF is very weak to track the mutation state; Strong tracking particle filter algorithm STF can improve the tracking ability of the mutation state, but it can't improve the particle degradation. In this paper, the STUPF algorithm is a good solution to solve these problems, and the simulation results also verify the effectiveness of the STUPF algorithm.