Image Optimisation for Hard-Field Tomography with an Irregular and Sparse Beam Array

J.L.Davidson1, N. Terzija1, C.A. Garcia-Stewart1, P. Wright1, K.B. Ozanyan1, S. Pegrum2,

T.J. Litt3, H. McCann1

1 School of Electrical Engineering, University of Manchester, UK.

2 Roush Technologies Ltd, Brentwood, UK.

3 AOS Technology Ltd, Melton Mowbray, UK.

Email : j.davidson-2@manchester.ac.uk

ABSTRACT

The application of Chemical Species Tomography (CST), using near-infrared absorption, inside the combustion chamber of an automotive engine generates two fundamental problems that are intimately linked: firstly, given the necessarily small number of path integral measurements through the subject, what is the optimum geometrical arrangement for image quality?; and secondly, which image reconstruction algorithm(s) should be used when the orientations of individual paths have no specific relationship to each other?

It has been found that an empirical approach to the array optimisation problem suggests an elegant general design rule in which the combined linear and angular coverage of the sinogram space is as uniform as possible, even at the expense of reducing the number of integral path measurements. This approach may also be adapted to provide a systematic way of coping with access constraints. In the work reported here, a maximum of only 27 measurements was possible in laboratory bench-top experiments with the array that was subsequently mounted into the engine.

To carry out practical phantom tests, a series of experiments was conducted with well-defined gas plumes issuing from flow-conditioning systems of known geometries. Median filtering was applied in the iterative Landweber algorithm to achieve robust image reconstruction of the above phantoms. Plumes as small as one-fifth of the subject diameter were reconstructed with good spatial fidelity.

Keywords image reconstruction, imager, hard-field tomography, limited view