Loading...
Evaluation and verification of five different image reconstruction algorithms for electrical resistance tomography applications
Author(s)
Deba, Charlie Nindjou
Date Issued
2016
Type
Thesis
Publisher
Cape Peninsula University of Technology
Abstract
Tomography is the ability to internally visualise an opaque medium or a body, using different imaging techniques. Electrical Resistance Tomography (ERT) technique is a method commonly used in process tomography. It uses a non-intrusive resistance measurement between a set of electrodes attached on the circumference of a fixed cross-section with a given conductivity and permittivity distribution. ERT appears to be simple, low cost, safe and non-invasive. Despite the advantages of ERT, the reconstruction of the internal conductivity of the pipe still face a crucial challenges such as noise, a relatively low spatial resolution, as well as ill-posedness of the inverse problem when doing the image reconstruction using reconstruction algorithms. Although previous work showed the potential of various algorithms for the reconstruction of ERT tomograms, no full characterisation and comparison of different algorithms could be found for real flow situations.
The ERT system was tested in the identification of different objects and fluid beds in a real time situation. The data collected from the measurements were then used for the image reconstruction using an algorithm developed by Time Long (One-step algorithm) and four EIDORS-based algorithms namely: Gauss-Newton algorithm with Laplace Prior (LP) and Gaussian prior (Automatic Hyper Parameter Selection (AHSP)), the Total Variation (TV) algorithm and the Conjugate Gradient (CG) algorithm. The performance of each algorithm was tested in different scenarios. The results obtained were then compared based on the quality and the accuracy of the images as well as the computational time of each algorithm. Firstly, reconstructed images were obtained using objects placed inside the ERT pipe test. Secondly, the algorithm performances were put to test in a level bed setup experiment and finally, the algorithm reconstructions were applied to the real flow situation, where different flow rates were applied. The results obtained were then analysed and compared.
The ERT system was tested in the identification of different objects and fluid beds in a real time situation. The data collected from the measurements were then used for the image reconstruction using an algorithm developed by Time Long (One-step algorithm) and four EIDORS-based algorithms namely: Gauss-Newton algorithm with Laplace Prior (LP) and Gaussian prior (Automatic Hyper Parameter Selection (AHSP)), the Total Variation (TV) algorithm and the Conjugate Gradient (CG) algorithm. The performance of each algorithm was tested in different scenarios. The results obtained were then compared based on the quality and the accuracy of the images as well as the computational time of each algorithm. Firstly, reconstructed images were obtained using objects placed inside the ERT pipe test. Secondly, the algorithm performances were put to test in a level bed setup experiment and finally, the algorithm reconstructions were applied to the real flow situation, where different flow rates were applied. The results obtained were then analysed and compared.
Additional information
Thesis (MTech (Electrical Engineering))--Cape Peninsula University of Technology, 2016.
File(s)![Thumbnail Image]()
Loading...
Name
207143528-Deba-Charlie-Nindjou-Mtech-Electrical-Engineering-Eng-2017.pdf
Description
Thesis
Size
3.4 MB
Format
Adobe PDF
Checksum
(MD5):d5f31fd0c4562ffd4581e1978ded8ccb