|
In this thesis, the image reconstruction for the electrical
impedance tomography (EIT) is implemented using the distributed
computing model based on the Winsock. It allows the standard
TCP/IP-based applications to be written for the Microsoft
Windows environment. The distributed computing model is built on
a cluster of local area networking PCs. It is also integrated
into our EIT system. The distributed computing model consists of
two parts: the client interface and the server interface. In
addition, the auto-scaling mechanism is introduced to solve the
loading unbalance problem. The workload is redistributed to each
server based upon their previous execution.Three algorithms
related to the image reconstruction are implemented using this
distributed computing model with the satisfactory performance.
In the finite element method based forward solver, which
includes 992 elements and 513 nodes, the speedup can be
increased to about 5.00 and the execution time can be decreased
from 54.82 seconds for one Pentium-120 based personal computer
to 14.41 seconds with the configuration of seven servers. With
the same configuration for the filtered backprojection
algorithm, the reconstructed image consisting of 101*101 pixels
and being mapped into the 992 triangular meshes, the speedup can
be increased to about 5.30 and the execution time can be
decreased from 18.62 seconds to 3.51 seconds. Similarly, in the
iterative scheme of Newton-Raphson method, the execution time of
10 iterations can be decreased from 534.09 seconds to
298.21seconds.
|