1990 Volume 15 Issue 44 Pages 47-54
This paper is the second report of our series of studies on the method to determine a velocity distribution from video images of visualized flow through image processing and numerical calculations. In the first report, we described the progress and background. the advantages and disadvantages of this technique, and the purpose of our studies. In this paper, we represent the outline of the system constructed in our laboratory, some examples of measurements, and the result of an experimental estimation of errors which arise in measurement. The system consists of a camera, a VTR, an image processor and a microcomputer. Flow images visualized with suspended tracer particles are once recorded on the VTR. The procedure to obtain velocity distribution is as follow; 1) An arbitrary selected one frame image is transmitted to the image processor. 2) The image is corrected to an image with uniform background, and subsequently it is reduced to a binary image. 3) The centroid coordinates and projected areas of tracer particles are calculated. 4) Other consecutive three frame images are treated simultaneously, and a set of data obtained from four frame images is transmitted to our university's computer, ACOS1000 or ACOS2000. 5) Velocity vectors are extracted by the tracer tracking method which was explained in our first report 6) Inadequate vectors are eliminated, where change of tracer area, tracer acceleration and velocity deviation are adopted as criteria for the judgment. A microcomputer version, in which inadequate vectors are eliminated manually, is also prepared. As examples of application a stationary vortex or a Karman vortex street behind a rectangular column in the water and an air flow in a ventilated room model were measured. As a tracer, styrofoam powders (for water) or metaldehyde aggregates (for air) were used. In every case, more than one hundred vectors are extracted from a set of data and resultant velocity distributions were considered to be reasonable. The error which arise in measurement was estimated through the velocity measurement of circular marks pasted on a turntable. The true velocity of a mark was calculated from rotating speed of the turntable and turning radius. The difference in declination between measured and calculated vector and the relative error of vector magnitude decreased as the mark displacement expanded. Hence, a measurement within 0.03 rad of declination error and within 5% of relative error of vector magnitude could be achieved provided that image sampling intervals are selected adequately.
