Abstract: For digital twin systems containing large-scale scenes,the rendering efficiency of the twin model has an important influence on the dynamic real-time performance and interactivity of the twin system.Aiming at the shortcomings of existing schemes in frustum culling and rendering frame,a new fast rendering method of digital twin model with fine-grained culling was proposed based on parallel computing framework.The model was divided into sub-models containing the same number of triangular slices,and the corresponding size of the bounding box was constructed according to the geometric coordinates of the sub-model.The fine-grained frustum culling reduced the scale of the model data in the rendering process.At the same time,the rendering process optimization and CPU/GPU collaborative processing were combined to build a parallel computing framework in the Graphics Processing Unit (GPU),and the frustum culling and calculation originally carried out in Central Processing Unit (CPU) were transplanted to the GPU to carry out the parallel calculation of frustum culling and operation,further reducing the computational burden on the CPU.The digital twin shop system of circuit breaker was taken as an example for verification.The verification results showed that the rendering time of CPU and GPU was reduced by 26.52% and 25.51% respectively,and the average frame rate of the system was increased by 28.35%,which effectively improved the operating efficiency and interactive performance of the digital twin system.

Key words: digital twin model, frustum culling, bounding box, large scale scenes, graphics processing unit

摘要： 对包含大规模场景的数字孪生系统而言,孪生模型的渲染效率对于孪生系统动态实时性和交互性具有重要影响。针对现有方案在视锥体剔除和渲染框架上的不足,提出一种基于并行计算框架的数字孪生模型视锥体细粒度剔除快速渲染新方法,将模型剖分为含相同数量三角面片的子模型,根据子模型几何坐标构建相应大小包围体,进行视锥体细粒度剔除,减少渲染流程中模型数据规模。同时,将渲染流程优化和CPU/GPU协同处理相结合,在GPU中构建并行计算框架,将原本CPU中进行的视锥体剔除和计算移植到GPU,进行视锥体剔除和操作的并行计算,进一步减少CPU计算负担。最后,以断路器数字孪生车间系统为例进行验证,结果表明,CPU、GPU渲染时间分别减少26.52%、25.51%,系统平均帧率提高28.35%,有效提升了数字孪生系统的运行效率和交互性能。

关键词: 数学孪生模型, 视锥体剔除, 包围体, 大规模场景, 图形处理器