Sammendrag
A new algorithm for fiber tracking in combined PIV/PTV measurement of fiber suspension flow is proposed based on SOM neural network and is examined by synthetic images of fibers showing 2D suspension flows. There is a new idea in the algorithm to take the orientation of fibers into account for matching as well as their position. In two-phase PIV measurements of fiber-laded suspension flows, fiber tracking has a key role together with PIV measurement of fluid phase. The essential parts of fiber tracking are to correctly identify and match fibers in successive images. The development of a method in order to determine the position and orientation of fibers using steerable filter with a reasonable accuracy have already been done, Carlsson et al. (2011). The present study is concentrated in the development of an algorithm for pairing fibers in consecutive images. The method used is based on the SOM neural network that finds most likely matching link in images on the basis of feature extraction and clustering. The fundamental concept is finding the corresponding fibers with the nearest characteristics, position and angle in images. It improves not only the robustness against loss-of-pair fibers between two image frames but also reliable matching at large numbers of dispersed fibers image using one more characteristics of fibers in image, namely their orientation, in addition to their coordinate vector.
Vis fullstendig beskrivelse
