Study Of PIV Parallel Computing Technology Based On GPU

Particle image velocimetry (PIV) is an important experimental method of fluid mechanics. But so far some of the research of the velocity of flow field are not fully realized in the real-time measurement, especially the cross-correlation which was did in the flow field images during analysising of processing is carried out in the CPU. Especially there are many pixels in the image, this will require a multi-threaded to handle, and in the multithreading the operation of the each point are the same, multithreading is repeated for each point of operation process. And in terms of the velocity vector diagram, the GPU is has the more advantage of its superior to the CPU, the development of the GPU is used for image rendering and display. The displaying of drawing which was did by CPU is only under GPU, it also slows down the speed of the whole system. So in this terms compared to the GPU CPU is no advantage.This PIV based on GPU parallel computing technology researching is made through the multithreaded operations on the pixel in the image. GPU can implement the tasks in multiple threads at the same time, it can greatly improve the processing speed of the whole system, combined with CUDA and OpenGL graphics drawing at the same time, also could speed up the graphics rendering and display of the system.The content of this experiment is seeding particles in the flow field, and then Shooting two frames during a very short time using CCD cameras, and making cross-correlation calculation on the both frames image, finally drawing the field velocity field. In this paper, the main content are the image cross-correlation calculation of two frames, and finally drawing the velocity field. This article mainly analyzes from the image matching parallel programming and the dynamic display of the reconstruction results.The parallel program design of image matching is mainly made in the Graphic Processing Unit (GPU for short). And the dynamic display of the reconstruction result is realized through the graphical programming interface OpenGL (English full name:Open Graphics Library). OpenGL is an open3d graphics software. The related content section later in this article will have detail.The experimental results show that the selection in the GPU cross-correlation analysis than in CPU can have a better drawing efficiency.