Real-time 3D imaging and display technology in virtual reality technology has become a key research object; However, modern computer work has been dramatically improved compared with the past; it still cannot meet the requirements of drawing realtime complex scenes and objects. Level of detail modeling technology is the most promising technology for 3D images, which is to develop a set of models with different levels of polygonal information from the same original model describing an event or object. Based on the above background, this article puts forward the application of 3D modelingbased LOD technology in virtual reality as the research goal, studies the concept of LOD technology in detail, analyzes the QEM algorithm, and then researches and optimizes the QEM algorithm. According to the experimental research in this article, the basic algorithm used in simplifying the initial model can simplify the model to a certain extent, but the data could be better. Using an immune genetic algorithm to facilitate nodes has succeeded dramatically, saving nearly 50% of nodes and about 90% of the time based on the most straightforward optimization. However, in this case, the model cannot display details.