期刊论文

hanmin

薛玉艳,qinpan,韩杰,姜长斌

2022-06-28

中国生物医学工程学报

电子信息与电气工程学部

J

2

190-197

0258-8021

The image reconstruction process is a typical ill-posed problem in magnetic induction tomography (MIT), in which the numerical solution is unstable. To solve this problem, an improved iteration Newton-Raphson algorithm based on weighted matrix and L<inf>1</inf>-norm regularization is improved. The proposed method adds the weight matrix in the objective function and adds L<inf>1</inf>-norm regularization term in L<inf>2</inf>-norm regularization penalty term. The analysis is made for three typical models in the data with and without noise, respectively. And the proposed algorithm is contrasted with Tikhonov regularization algorithm and iterative NR algorithm. In the data without noise, relative to Tikhonov regularization algorithm and iterative NR algorithm, the relative error is reduced by 0. 11-0. 14. And then, the correlation coefficient is raised by 13% - 17%. The algorithm has good performance in imaging. In the data with noise, the relative error is reduced by 0. 06-0. 09, and the correlation coefficient is raised by 7% - 10% in the proposed algorithm. The algorithm has good anti-noise performance, which has offered theory basis for the study of reconstruction accuracy.

新增回溯数据