<p>In contrast to a conventional four-stroke engine, the opposed-piston two-stroke hydrogen internal combustion engine not only delivers superior power density and reduced heat rejection, but its high volumetric power output also mitigates drawbacks, such as poor emissions performance and fuel short-circuiting, making it highly significant for engineering research. However, the unique fuel properties of hydrogen combined with the complex architecture of the opposed-piston two-stroke internal combustion engine result in the simultaneous occurrence of fresh charge losses during both the scavenging and exhaust phases. This dual loss mechanism prevents the engine from achieving its ideal air-fuel ratio. To address this issue, this paper proposes an Extended Kalman Filter algorithm driven by both data and mechanisms to achieve accurate online estimation of in-cylinder gas mass. First, by analyzing the thermodynamic processes and gas flow dynamics within the cylinder during the intake and exhaust phases, a state-space equation for the exhaust and scavenging process is established, for the foundation for subsequent in-cylinder mass estimation. Second, data collected during the exhaust flow process are used to fit the complex exhaust flow equation, thereby reducing the strong coupling, time-varying, and non-linear characteristics of the state-space equation. Finally, to overcome the limitations of the static noise covariance matrix and improve the robustness and accuracy of the state estimation of the Extended Kalman Filter in dynamic systems, the Data-Mechanism Extended Kalman Filter (DM-EKF) is proposed. The fitted exhaust flow rate is introduced in the state prediction and update steps of the DM-EKF to dynamically correct the process noise covariance. Through simulation experiments in Simulink, the experimental results demonstrate that the proposed method has good performance in estimating the state variables during the exhaust scavenging process.</p>