Traditional simulation technologies are limited by algorithm complexity, computing power and interface bottlenecks. To solve this problem, a system which implements real-time dynamic infrared scene simulation in real images on the basis of augmented reality is designed. Firstly, 3DS MAX is used to establish a virtual target model. The geometric consistency transformation and infrared radiation consistency transformation are carried out for the virtual target model. Then, SURF-ORB feature matching is used to track the motion of the detector. Finally, the virtual target is fused with the real scene through raster rending. Thus, the design of the dynamic infrared scene simulation system based on augmented reality is completed. The experimental results show that high frame rate and high realism infrared simulation images can be obtained by this system. Its scalability is very strong.