Abstract：In the field of Electromagnetic Tomography (EMT), solving inverse problem of ill-conditioned matrix equation is an important step in image reconstruction. This paper focuses on the parameter selection in the EMT inverse problem of the ill-conditioned matrix equation with Truncated Singular Value Decomposition (TSVD) algorithm and Tikhonov regularization method. It presents a new parameter selection algorithm based on maximizing the correlation coefficients of the sample image and its feasibility is analyzed in contrast with the original parameter selection algorithm. The multi-sample characteristics of the sensitivity matrix presented in this paper are applied to the algorithms of L-curve and generalized cross-validation. The statistical methods are used to improve the stability of the sensitivity matrix equation and to prevent accidental errors caused by single sample solution to a certain extent. Finally, a hardware imaging platform is built to demonstrate the actual imaging effect of the proposed algorithm. The advantages and disadvantages of singular value restraining mode and interception mode in the practical application and parameter calculation of EMT inverse problem are also briefly discussed.The experimental results show that the proposed algorithm has better imaging effect in the test samples.