The Unscented Kalman Filter is well-suited for processing highly nonlinear dynamic systems and is a nonlinear Gaussian filter. However, due to the high nonlinearity of the systems, the problem of non-local effects caused by the offset of the state has not been effectively solved. A smaller step can improve estimate accuracy, but it will increase the computational load, so a trade-off needs to be made when selecting the step. This paper proposes an adaptive micro-step UKF algorithm that relies on the symmetry of sigma points. The proposed method reduces the error of state prediction by incorporating micro-steps in highly nonlinear interval of the systems. A criterion based on symmetric sigma points is developed as a guideline to decide whether to start the micro-step UKF algorithm. The performance of the proposed method is illustrated through a re-entry trajectory target tracking model. Simulation results indicate that, under the conditions of the system is highly nonlinear and the measurement frequency is limited, the proposed method outperforms several reference algorithms in precision.