Subway tunnel flooding is one of the typical scenarios in urban waterlogging. When the emergency decision-making is untimely and inaccurate, it can lead to serious casualties. However, it is expected that through some responsive monitoring methods, an intelligent system can be constructed to achieve the real-time perception of the underground flood situation and to predict the process and intensity of the disaster. This study is based on a scale model experiment of flooding in straight subway tunnels. Firstly, it classified the basic patterns of the flood process. Some fitting formulas were respectively established for water depth and kinematic parameters. A calculation method for quickly predicting the flood inundation process was constructed. Secondly, this paper proposed a mode for the monitoring and early warning system. Based on the flood patterns and characteristic variables in subway tunnel flooding, the system can quickly determine the flow rate and entry position of flood and deduce the flood trend in the future at a relatively low cost. The system workflow was divided into three phases, and the mechanism for each phase of the system was sorted out. It could help to provide scientific and reliable information for emergency decision-making and gain time for the emergency response.