Based on the principle of compressed sensing,a method of constructing nonconvex entropy(NE)function as regularization term is proposed,which can effectively alleviate the inverse problem of electrical capacitance tomography(ECT)ill-condition and ensure the sparseness of the solution.Fast iterative threshold contraction algorithm(FISTA)is used to accelerate the convergence rate.The obtained solution is optimized by Gaussian mixture model(GMM),and the model parameters are updated by expectation maximization algorithm(E-M).After that,NE-GMM algorithm is obtained.Both simulation and experimental results show that reconstructed images with the best quality can be obtained using NE-GMM algorithm compared with LBP,Landweber,iterative hard threshold(IHT),ADMM-L1 and NE algorithms,especially the fidelity of center distribution and multi-object distribution is further improved.The average relative error and correlation coefficient of the simulated reconstructed image obtained by this method are respectively 0.4611 and 0.8827,which are superior to the other five methods.
image reconstructionelectrical capacitance tomographynonconvex entropyGaussian mixture model
国家科技期刊平台
NSTL
万方数据
