Abstract
In order to measure the velocity distributions in a tank using VTR images, a new image processing algorithm was developed, in which the expert system was used. The histogram of candidate velocity vectors was constructed. The correct vector density became larger than the false vector density in the velocity vector histogram, consequently the most probable vector was defined. The histogram was re-constructed using the information of the probable vector in order to reduce the noise and to improve the accuracy of the velocity data. For this re-construction, the method of the expert system is introduced in order to select the feed back parameters. Using this algorithm, the velocity fields could be obtained even from noisy images. The present algorithm was applied to the two-dimensional flows in a thin rectangular tank with free surface and the three-dimensional flows in a open channel, showing high effectiveness and high extensiveness.
