Image Reconstruction Algorithms for High-Speed Chemical Species Tomography

C A Garcia-Stewart, N Polydorides, K B Ozanyan and H McCann

Department of Electrical Engineering & Electronics, UMIST, Manchester, UK, h.mccann@umist.ac.uk (All authors are members of the Virtual Centre for Industrial Process Tomography)

ABSTRACT

High-speed tomography requires the use of simultaneous measurement and no moving parts. This study describes two methods developed to generate smooth images of hydrocarbon gas distribution from a limited data tomography system with 32 measurements. Both methods include the use of a- priori knowledge of the gas cloud’s smooth variation in concentration. Each method performs interpolation either prior to or after the inverse problem calculation. The first technique uses the sensitivity map in the Landweber iteration method to generate a space image and then implements a weighted box averaging method. The sparse point image is then interpolated to generate a smooth image. The second method implements interpolation of the sinogram raw data, interpolating angularly and spatially. The newly generated data is inverted by the Filter Back Projection (FBP) method to generate an image. A forward calculation is made of the image generated and Landweber-like iterations are performed. Both techniques generate acceptable images from experimental data.

Keywords Limited data reconstruction, sinogram, tomography, a priori information