Authors: Hannes Petermeier Wojciech Kowalczyk Antonio Delgado Cornelia Denz Frank Holtmann
Publish Date: 2007/02/21
Volume: 42, Issue: 4, Pages: 611-623
Abstract
In biological fluid mechanics powerful imaging methods for flow analysis are required for making progress towards a better understanding of natural phenomena being optimised in the course of evolution At the same time it is of crucial importance that the measuring and flow visualisation techniques employed guarantee biocompatibility ie they do not distort the behaviour of biosystems Unfortunately this restricts seriously the measures for optimising the image generation in comparison to other flow fields in which no biological systems are present As a consequence images of lower quality leading to erroneous artefacts are obtained Thus either novel detection techniques that are able to overcome these disadvantages or advanced evaluation methods enabling the sophisticated analysis and description of flow fields are essential In the present contribution both areas are covered A novel socalled neuronumerical hybrid allows to detect artefacts in conventional experimental particle image velocimetry PIV data of microorganismic flow fields generated by ciliates The handling of artefacts is performed by the hybrid using a priori knowledge of the flow physics formulated in numerical expressions and the enormous potential of artificial neural networks in predicting artefacts and correcting them In fact the neuronumerical hybrid based on the physical knowledge provided by the Taylor’s hypothesis can detect not only spurious velocity vectors but also additional phenomena like a moving boundary in the present case caused by the contraction of the zooid of a microorganism Apart from the detection of the artefacts a correction of the spurious velocity vectors is possible Furthermore a method to detect microscopic velocity fields based on nonlinear optical filtering optical novelty filter ONF is presented On the one hand it can be employed to expose phase changes in flow fields directly from the nonlinear response and without additional tracers On the other hand it can be used to preprocess low quality images of flow fields loaded with particles and extract the motion of particles with an enhanced contrast The flow fields obtained by the correlation based PIV method of the ONF filtered and unfiltered image sequences are compared and discussed
Keywords: