A partial sampling strategy was recently proposed to make the computational complexity of the unscented Kalman filter (UKF) quadratic with the state-vector dimension. However, the quadratic complexity remains a thorny issue in the large SLAM. To solve this problem, this paper presents a filtering solution for the SLAM problem called shrink unscented Kalman filter (S-UKF). It firstly proves that equivalence of the whole and partial sampling strategy for the decoupled nonlinear systems. Then a shrink form is presented by reconstruction the cross-correlation items to reduce the computational complexity. Finally, the simulation results and experimental results based on real environmental data sets validate the effectiveness of this method.