Image reconstruction of electrical capacitance tomography based on non-convex and nonseparable regularization algorithm
Two-phase mixing in a stirrer is a common phenomenon in chemical production.Electrical capacitance tomography(ECT)technology mainly visually reconstructs the distribution of the two phases for monitoring purposes.Inspired by sparse Bayesian learning,a non-convex and nonseparable regularization(NNR)algorithm is proposed to reconstruct ECT images.The low-rank characteristics of the matrix are introduced on the basis of the sparse prior,and a new optimization problem is proposed in the latent space by using the maximum posterior estimation.Dual variables are used to map the objective function of the latent space to the original space for an iterative solution,which is used to restore the simultaneous sparse and low-rank matrices.Compared with the convex approximation L1 norm,the NNR algorithm can obtain more accurate reconstruction images,and it is easier to converge to the global optimal solution than the non-convex separable method.To verify the reconstruction effect of the NNR algorithm,the reconstruction was compared with the other five algorithms through numerical simulation and static experiments.The results show that the NNR algorithm can effectively reduce reconstruction artifacts,improve the reconstruction quality of the central object,and provide a high-quality reconstruction algorithm for the two-phase distribution in the stirrer.
electrical capacitance tomographyimage reconstructionnon-convex and nonseparable regularizationsparse-low-rank modeltwo-phase mixture
国家科技期刊平台
NSTL
万方数据
