A building fire prediction and evacuation system based on cellular automata is proposed for densely populated large buildings. Firstly, the grid method is used to establish a planar model about the building, and the cellular automaton model is applied to realize the fire prediction and the crowd evacuation on the same interface. Then, the best exit design scheme for evacuating the crowd is obtained by changing different parameters of the exit. Finally, the improved ant colony algorithm is adopted to optimize the evacuation lights in the building, so that it can dynamically indicate the direction of escape. The simulation results show that the system can effectively solve the problem of exit design and personnel safety evacuation in the event of fire in the building group.