Non-linear Image Reconstruction Technique on Electrical Capacitance Volume Tomography (ECVT) Using Combined Dual Neural Networks Approach

    Pamuji Widodo, BS Thesis, Dept. of Physics, University of Indonesia, 2009

    Radial basis network (RBN) and analogue Hopfield network is developed for nonlinear image reconstruction of electrical capacitance volume tomography (ECVT). The (nonlinear) forward problem in ECVT is solved using the RBN trained with a set of capacitance data from measurements. This RBN is able to apply rather many neurons than standart feed-forward. The inverse problem is solved using an analogue Hopfield network based on a neural-network multi-criteria optimization image reconstruction technique (HN-MOIRT) and algebraic reconstruction technique (ART) formed from forward solution RBN. The nonlinear image reconstruction based on this HN-MOIRT approach is tested on measured capacitance data not used in training to reconstruct the permittivity distribution. Whereas on This ART uses partial differncial technique which integrated by fNN of training result of RBN. The performance of the technique is compared against commonly used linear Landweber, semi-nonlinear, and nonlinear image reconstruction techniques, showing superiority in terms of both stability and quality of reconstructed images. Image reconstruction process was conducted using  MATLAB R2007b software.

    Keywords:
    ECVT, RBN, HN-MOIRT, ART, optimization technique, Image reconstruction, MATLAB R2007b.