Pose tracking is one of the most critical techniques in mobile robot navigation, but dynamic environments will greatly reduce its accuracy and robustness. For addressing the above problem, this paper proposes a pose tracking method based on Monte Carlo localization and Entropy-based trimICP (E-trimICP) to achieve accurate and robust robot localization. First, a hybrid noise filtering method that fuses distance filtering and radius filtering is presented to improve the reliability of observation data. Then, an E-trimICP is designed to improve the accuracy of scan matching in dynamic environments. The proposed method can not only enhance the adaptability by determining the trim ratio based on the dynamic degree of the environment, but also improve the operational efficiency through the improved heap sorting. Ultimately, the effectiveness of the proposed pose tracking method is verified by real experiments using a self-developed mobile robot.