In recent years, AR (Augmented Reality) technology has attracted extensive attention from researchers. Driven by computer vision and artificial intelligence technology, AR technology has shown a strong momentum of development. Both the display equipment performance, tracking registration accuracy and the nature of human-computer interaction have been greatly improved. However, it can be seen that there are still many problems to be solved in augmented reality technology. In the case of tracking registration technology, the current tracking registration technology can only make use of a small amount of information in the scene, such as feature point information, which leads to an incomplete understanding of the system to the environment.