전기 임피던스 단층촬영법에서 RANSAC 알고리즘을 이용한 고장 전극 검출

Abstract

Electrical impedance tomography(EIT) is useful imaging technique which reconstructs internal resistivity distribution based on injected currents and measured voltages. In EIT, to get satisfactory reconstruction results, it is important to use reliable current-voltage data as well as good inverse algorithm. Therefore, various methods have been developed for checking EIT system. The main objective of existing methods is to evaluate and optimize the EIT system to obtain reliable measurement data. However, most methods need extra cost and time for test. In this paper, modified Gauss-Newton method based on random sample consensus(RANSAC) is proposed to reconstruct the internal resistivity distribution with trustworthy measurement data. Some incorrect data are extracted from measurements and they are excluded in the reconstructions. In the modified Gauss-Newton method, residual error equation and sampling scheme are proposed to detect fault electrode. To evaluate the performance of the proposed method, numerical simulation and experiment are performed and reconstructed results are compared with conventional Gauss-Newton method. From reconstructed results, the proposed method shows good performance in the reconstructions.