As to the shortcomings of the temperature field modeling based on the thermodynamics method in the infrared signature scene simulation, the BP neural network is brought into the infrared temperature field modeling of objects in the scene, and weights of the neural network are optimized by the genetic algorithms. Through frequent testing on the temperature of the objects, the training sample assemblies are obtained. Then with the neural network, the temperature field modeling of the scene consisted of car and cement road surface is built and the infrared images of the scene of the day and the night under different set weather conditions are simulated. According to the analysis of the results of the simulation, it′s proved that this model can simulate the infrared images of the scene under set weather conditions more accurately in real time.