To settle the pedestrian evacuation problems of large public building, by taking advantages of Ant Colony Algorithm in solving the shortest path problem while taking the behavioral characteristics of persons during fire into consideration, this paper puts forward a dynamic optimizing model of person evacuation path based on Ant Colony Algorithm, and combines with the analogue simulation of fire-free or fire occurrence examples, finally goes to a conclusion that dynamic adjustment of the optimized evacuation path can improve the evacuation efficiently and reduce the casualties effectively.