摘要
流血模拟是虚拟手术训练系统的重要组成部分。然而,流血模拟的真实性和庞大的计算量对手术训练系统具有很大的挑战。提出一种基于GPU加速的方法以解决虚拟手术中的流血效果模拟的实时性问题。该方法利用网格法实时划分问题区域,创建以支持域为边长的空间网格。通过临近网格搜索最近相邻粒子,并且通过并行计算架构(CUDA)多线程并行加速技术完成粒子控制方程的求解以及血液与固体交互的计算,大大提高了运算效率,从而提高了手术训练的实时性。一种改进的移动立方体算法(marching cube)用于流体表面的渲染,大大提高了手术训练的真实性。实验结果表明,该方法能够满足虚拟手术过程中流血的模拟需求,相比于CPU的实现,速度明显提升。
Blood simulation is an important part in the virtual surgery training system. However, the huge computational complexity and authenticity of blood simulation is of great challenge to the surgical training system. In this paper,a simulation method based on GPU-accelerated was used for blood simulation in surgical training system. The grid method was used to divide the target area,create space grid domain,and search neighboring particles by neighboring grid. We solved the particle control equation and calculated the interaction between blood and solid by parallel computing architecture( CUDA) multi-threaded parallel acceleration technology,which greatly improved the operational efficiency and improved the real-time of training. In addition,an improved marching cube algorithm was used to render the surface of fluid, which improved the authenticity of surgical training. Experimental results showed that the authenticity and flexibility of blood met the simulation requirements during the surgical training when using our method.Furthermore,the speed of blood simulation was significantly improved comparing to the realization of CPU.
出处
《计算机应用与软件》
北大核心
2018年第2期242-247,315,共7页
Computer Applications and Software
