To address the accuracy and stability issues in underwater target tracking, it is essential to consider the impact of nonlinear state as well as measurement equations on filtering performance, as well as the potential adverse effects of noise and other interfering factors that can render the covariance matrix coefficients non-positive-definite. In this study, we use the Singular Value Decomposition-based Unscented Kalman Filter (SVD-UKF), along with the application of Rauch-Tung-Striebel (RTS) smoothing for tracking the target's position. Three filtering algorithms are evaluated for their tracking performance through simulated analysis: the Unscented Kalman Filter (UKF), SVD-UKF, and RTS-UKF. The results indicate that, in the presence of nonlinear state and measurement equations, applying RTS smoothing with coefficient adjustments and utilizing SVD-UKF leads to higher filtering accuracy in both position and velocity estimation. The improved algorithm exhibits a 40% reduction in the Root Mean Square Error of position compared to the standard UKF. Furthermore, the filtered tracking trajectory demonstrates enhanced stability.