A New Index for Hyperparameters Selection in ERT Reconstruction Algorithms

As an advanced visualization method, Electrical Resistance Tomography (ERT) reconstruction algorithms have increasingly been focused on. High-quality ERT reconstruction algorithms generally involve hyperparameters, and selecting the optimal ones is vital for achieving high reconstruction quality. However, in practice, selecting the optimal hyperparameters is a challenging thing due to the unknown ground truth. In this paper, a new evaluation index for selecting hyperparameters in ERT reconstruction algorithms is creatively proposed. The new index is based on a nonlinear physical equation which can measure the difference between the reconstruction results and the unknown ground truth. Results from simulation and static experiments using two representative reconstruction algorithms demonstrate that the proposed index is both valid and more precise. The proposed method provides a theoretical basis for further improving the ERT reconstruction quality.