Test image of 512x512 pixels containing 1024 particles. 

Context in source publication

Context 1

... theory of the HRPV procedure was described and evaluated by Keane et al. [ 11]. In an experiment of analysing turbulent flow PIV data they found an increase of spatial resolution by a factor of 2.5 in each direction. Furthermore Cowen and Monismith [6] employed the technique with synthesised images and for images recorded from a turbulent boundary layer. They conclude that the results of the application of the HRPV method are superior to the results attained by PIV alone. This in the sense that not only resolution can be improved but also that the accuracy can be enlarged. The difference of the present method with the two mentioned above is in the particle tracking algorithm. Both Keane et al. [11] and Cowen and Monismith [6] use a window method defining the estimated area to which a particle advects. If there are two particles in this window the matching is ambiguous and must therefore be discarded. Obviously no matching can be established when there is no particle at all in this window. By minimising the global constraint, as specified in the matching algorithm section 2.5, the present PTV method should be able to obtain a larger yield without loss of accuracy. Thus the advantages of using HRPV compared to PIV are the enlarged yield and accuracy. With respect to PTV the advantages are the ability to process only two sequential images at relatively large image densities and image sequences $! with #" relative large advection distances. This can be done with only a small maximum matching distance, , obtaining high quality data. If we are not dealing with a starting flow, the PTV algorithm as described in previous sections shows a transient in the quality of the output. Errors could be kept low by using a small maximum matching distance, but then the yield will show a transient starting at very low values to higher values due to the neighbourhood estimation. However this will work only if the seeding density per flow structure is high enough, a condition that also has to be met for performing a successful PIV. Thus, if the seeding density per flow structure is large enough an additional advantage is obtained by the application of HRPV by omitting the transients. The HRPV algorithm is implemented by considering the background displacements estimated by PIV as additional matchings in the neighbourhood estimation. By omitting the temporal extrapolation we obtain a separated scheme for PIV estimation and PTV matching. Involving also the temporal extrapolation results in a mixed scheme in which only the displacement of previously unmatched particles is estimated by both the surrounding matchings of PTV as well as the PIV displacements. Besides performing a PIV estimation for the entire sequence the implementation can handle the use of PIV estimation of one image pair for the entire sequence in case of steady flow or statistically steady turbulent flow (with relatively low turbulence intensity). Furthermore the present HRPV processing is also able to handle sequences in which there is only a random set of PIV estimations of all possible subsequent image pairs. The developed algorithm was tested with synthetic images of a well defined flow field. The synthetic images were composed of several components representing important features of real image sequences. The images are grey level images with 256 grey values, 0 to 255. Particles are represented as Gaussian shaped intensity blobs. For each pixel value the particle intensity is integrated over the pixel area. An example of a test image is shown in figure 4. High wave number noise as well as background variations can be added easily. The stream function describing the 2D vortical flow field chosen for advecting the randomly ...