Reconstruction Limitations for Electrical Resistance Tomography

R K Y Chin, S C Murphy, TAYork

Sensing, Imaging and Signal Processing Group, School of Electrical and Electronics Engineering, University of Manchester, t.york@manchester.ac.uk

ABSTRACT

In industrial applications of Electrical Resistance Tomography (ERT), reconstructed images are often used to validate the performance of an ERT system. This practice is often subjective and lacks robustness. Importantly, no numerical results that map the true performance of an ERT system across the imaged space can be extracted. This paper aims to promote efforts to adopt a more scientific and comprehensive approach to evaluate various contributory parameters that dictate the quality of a reconstructed image. It considers the issues that inform such measures and focuses on two analysis techniques that have been used to compare the quality of measurements. Firstly, spectral analysis using Singular Value Decomposition (SVD) is used to extract information on the quality of the measurements. The Discrete Picard Condition (DPC) is applied to determine the stability of the dataset. Secondly, resolution matrixes are used to analyse spatial resolution across the imaged space. The effect of the varying the level of regularization on the resulting stability of measurements due to measurement noise is considered.

Keywords Electrical Resistance Tomography, image reconstruction, spatial resolution, 3D measurement strategy