Modern virtual reality technology shows great application potential in the fields of education and training, industrial design, medical simulation and so on. Aiming at the problems of low efficiency of dynamic scene construction and insufficient visual expression in virtual simulation environment, this paper proposes a game engine-driven optimization strategy based on virtual simulation environment. The methodology adopts Unreal Engine 5.2 to construct the virtual scene geometric modeling, uses the QuadriFlow mesh reconstruction technique of shielding Poisson’s equation to realize the model lightweight, combines the improved visual SLAM algorithm for dynamic feature point removal, and improves the scene visual dynamic expressiveness with the aid of the LK optical flow method. The experimental results show that the optimized virtual scene rendering speed reaches 7.865ms, 5.917ms and 8.653ms respectively, the indexing efficiency is improved by about 65% compared with the traditional method, the memory consumption is reduced by more than 55%, and the average value of the SSIM score reaches 0.9756.The method effectively solves the technical problems of dynamic scene construction in the virtual simulation environment, and provides reliable technical support for high-quality virtual reality application provides a reliable technical support, and significantly improves the visual performance effect while ensuring the real-time performance.
