Study On Algorithms Of Particle Image Velocimetry

Particle Image Velocimetry (PIV) and its image measurement techniques have been rapidly developed with the drastic development of image processing method and computers. As a tool for measuring the whole instantaneous flow field without contacting the flow, PIV and related image measurement techniques have been extensively applied to automated measurements of multiphase flows, turbulence and thermal flows with reliable accuracy. Related techniques and their applications are briefly reviewed in this paper, the conventional cross-correlation algorithm and the cross-correlation algorithm based on Fast Fourier Transform (FFT) are discussed. The later is fast and makes the real-time of PIV possible.However, due to the limitation of Fourier Transform, the simultaneous improvement of calculation accuracy and speed are limited. Fourier Transform cannot offer entire information of signals, namely frequency information presenting to any period of time cannot be confirmed on earth. Although, Short Time Fourier Transform can obtain local information, it is difficult to find a "good" time window to fit for every period of time.To conquer the limitation of Fourier Transform, an image processing method based on wavelet transform and multi-resolution analysis ispresented in this paper. The most excellence of wavelets transform is the multi- resolution analysis. Wavelets transform has localized character and self-acclimated function. The particle images are processed directly in spatial domain for this peculiarity, Two vision paths are simulated by wavelet multi-resolution decompositions of two-dimensional images, and matching calculations are carried through in each path. In this way, a fast matching model is founded.The accuracy of calculation is improved for direct calculation in spatial domain, and the speed of calculation is faster than the conventional cross-correlation algorithm for eradicating plenty of illusive aims to eliminate the redundant operations farthest.Finally, the realization procedures of the method via VISUAL FORTRAN are provided. The computational results of the particle images are agreement with the simulated datain reasonable, the most absolute difference of the displacement is 0. 6671 pixel at the X abscissa and 0. 7928 pixel at Y abscissa; the computational results are equal to the data form the algorithm of particle brightness-distribution pattern tracking (the conventional cross-correlation algorithm) at voluminous points, their discrepancy is only one pixel at few points, mostly in boundary area. The algorithm via wavelets transform and multi-resolution analysis can economize about 38.33 percent in computational time. It is applied to the actual particle images and bubble image for the flow in a plunge pool. The computational results are satisfied. It is applied to the actual particle images and bubble image for the flow in a plunge pool. The computational results are satisfied.