An iterative thresholding algorithm for the inverse problem of
electrical resistance tomography
$
Lingling Zhang
a
, Genqi Xu
a
, Qian Xue
b,
n
, Huaxiang Wang
c
, Yanbin Xu
c
a
Department of Mathematics, Tianjin university, Tianjin 300072, PR China
b
College of Aeronautical Automation, Civil Aviation University of China, Tianjin, 300300, PR China
c
School of Electrical Engineering and Automation, Tianjin university, Tianjin 300072, PR China
article info
Article history:
Received 30 January 2013
Received in revised form
27 May 2013
Accepted 22 July 2013
Available online 9 August 2013
Keywords:
Electrical resistance tomography
Sparsity
Iterative
Surrogate term
Regularization
abstract
Image reconstruction in Electrical Resistance Tomography (ERT) is an ill-posed nonlinear inverse
problem. Considering the sparsity property of ERT model, in this paper, we replace the conventional l
2
regularization penalty term by weighted l
p
ð1r po 2Þ penalty term. To overcome the non-quadratic
property, a surrogate term is added to the objective function. An interesting condition is that the classical
methods (e.g. SVD, Landweber iteration) can be used to solve the l
p
ð1r po 2Þ least squares problems.
Both typical and complicated distributions (e.g. annular and cross-shape) have been examined using a
16-electrode configuration based on the finite element method (FEM) software COMSOL. The simulated
results demonstrate the feasibility of the proposed algorithm, and compared to the l
2
regularization
method, the proposed algorithms can produce images of higher quality, which are evaluated both
qualitatively and quantitatively.
& 2013 Elsevier Ltd. All rights reserved.
1. Introduction
Electrical tomography (ET) is a widely used imaging technique
which intents to estimate the interior distribution from boundary
measurements [1,2]. Due to the advantages of high speed, safety,
non-invasiveness and low-cost, ET has made a great progress over
two decades and has been widely applied in the fields of
petroleum, chemical, astronomy, geology survey, non-destructive
testing etc. For electrical resistance tomography (ERT), multiple
electrodes are arranged around the boundary of the vessel at fixed
locations in such a way that they make electrical contact with the
fluid inside the vessel but do not affect the flow or movement of
materials. In principle, ERT can be used to investigate and monitor
any process where the main continuous phase is at least slightly
conducting and the other phases and components have differing
values of conductivity. A typical application is real time monitoring
of multiphase flows within process engineering units. Specific
applications where ERT has been successfully exploited include
solid/liquid and liquid/gas mixing, hydrocyclones, packed columns,
flotation columns, precipitation processes, liquid-liquid extraction
and hydraulic conveying [3– 5].
The major challenge in ERT is the inverse problem, namely the
image reconstruction problem, due to its nonlinearity and ill-
posedness [6–8]. To address this problem, many approaches have
been proposed, e.g. conjugate gradient method, Landweber
method, Tikhonov regularization method, singular value decom-
position (SVD) and truncated singular value decomposition (TSVD)
[9–11]. These optimization methods are effective for recovering
smooth signals. However the signals to be reconstructed in ERT are
sparse. The reconstruction images obtained using the aforemen-
tioned methods are often over-smoothed and cannot meet the
requirement to image quality in many cases.
With the development of the compress sensing theory [12,13],
an l
1
regularization method, which attempts to find the solution of
a sparse linear inverse problem has attracted considerable atten-
tion in the signal processing literature [14–17], and various l
1
regularization algorithms have been applied to image reconstruc-
tion in electrical tomography. Various experimental results
demonstrate that the l
1
regularization algorithms are superior to
the l
2
regularization methods in the imaging quality. The only
drawback of these methods is the relatively poor real time
performance [18–20].
In this paper, we apply the l
p
ð1r po 2Þ regularization method
to the ERT inverse problem. Due to the utilization of non-quadratic
constraints which bring nonlinearity to the method, it is difficult
to obtain an ideal result with high speed. To solve this problem, a
surrogate term is added to the obje ctive function. The resulting
optimization problem can then be solved by using a Landweber -type
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/flowmeasinst
Flow Measurement and Instrumentation
0955-5986/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.flowmeasinst.2013.07.010
☆
This document is a collaborative effort.
n
Corresponding author. Tel.: 86 022 27405724.
E-mail addresses: xueqian1987@163.com, xueqian@tju.edu.cn (Q. Xue).
Flow Measurement and Instrumentation 33 (2013) 244–250