Subway emergency may lead to passengers' panic, especially under self-organizing circumstance, which will spread rapidly and have an adverse impact on the society. This paper builds an improved SIRS model of passengers' panic spread in subway emergency with consideration of passengers' density, the characteristic of subway car with the confined space, and passengers' psychological factors. The spread of passengers' panic is simulated by use of Matlab, which draws the rules of how group panic spreads dynamically. The trend of stable point of the infection ratio is analyzed by changing different parameters, which help to draw a conclusion that immunization rate, spontaneous immune loss rate, and passenger number have a great influence on the final infected ratio. Finally, we propose an integrated control strategy and find the peak of passengers' panic and the final infected ratio is greatly improved through the numerical simulation. The research plays a vital role in helping the government and subway administration to master the panic spread mechanism and reduce the panic spread by improving measures and also provides certain reference significance for rail system construction, emergency contingency plans, and the construction and implementation of emergency response system.