In order to solve the problem of low drawing accuracy of complex scenes in virtual reality system, the research on improving the modeling accuracy of digital environment driven by intelligent technology is put forward. Firstly, the half-edge folding algorithm is used to simplify the complex three-dimensional geometric model, which reduces the computational overhead and improves the efficiency of model rendering. Then, in the process of drawing the level of detail of complex scenes, the preprocessing stage of selecting the level of detail is added, and the appropriate parallax measurement coefficient of the level of detail is selected for each frame of scene area through the calculation of projection value, which speeds up the real-time calculation. The experimental results show that the number of display triangles of the optimized method in this paper is increased, that is, the time complexity is improved. However, due to the improved preprocessing mechanism of LOD model selection, the total execution time of the algorithm is still less than that of SR-LOD method, which is 31.145s, that is, the complex scene can still be drawn relatively quickly.