Abstract: Virtual simulation of fluid phenomena is of great value, and SPH fluid simulation-based method can truly reflect the motion of the fluid. However, as the size of the fluid increasing, the computation becomes large, especially fluid collision detection in complex scenes significantly reduces the efficiency of fluid simulation. By using GPU to accelerate SPH physical calculation and collision detection, and use of new neighbors particles chain table building method for particles searching, can improve fluid simulation efficiency. By using voxel scene rasterization of sampling methods, the calculation of collision detection was speeded up, and the storage space was saved by compressing each voxel value into 1bit and saving it to file. Preliminary experiment results show that the method is efficient for collision detection processing, and can reduce processing time of collision detection.

Key words: SPH, GPU acceleration, collision detection, fluid simulation, voxelization