In the rocket projectile firing assessment test, the measurement data from a ground-based tracking radar may have poor quality due to complex interferences, especially in the rocket projectile’s low-altitude flight phase corresponding to the radar’s low-elevation detection that is corrupted by ground clutter. In this work, we are concerned about the problem that large errors accompany the rocket projectile’s trajectory that is reconstructed with polluted radar measurements, and affect the accuracy of assessment. We present a simple but practical robust smoothing method that combines robust local weighted regression smoothing (RLWRS) and quartile mean smoothing (QMS). Concretely, we use the RLWRS method to smooth the reconstructed sub-trajectory in the radar’s low-elevation detecting phase, whereas we use the QMS method to smooth the rest. Finally, a relatively realistic trajectory of the rocket projectile is obtained by splicing these smoothed sub-trajectories. A simulation experiment is provided and performed to verify the effectiveness of the RLWRS-QMS method.