Improvement of image reconstruction in electrical capacitance tomography (ECT) by sectorial sensitivity matrix using a K-means clustering algorithm

A sectorial sensitivity matrix using a K-means clustering algorithm has been proposed for electrical capacitance tomography (ECT) for accurate determination of permittivity distribution in the inclusion within the background area. The proposed sectorial sensitivity matrix clusters the pixel position of the inclusion area to minimize the soft field effect that leads to image artifacts and blurred images. The standard sensitivity matrix is nonlinear thus it is not easy to eliminate image artifacts due to the noise signal and nonlinear electric field propagation. The performance of the sectorial sensitivity matrix is evaluated through numerical simulation and experimental studies. From the results, the sectorial sensitivity matrix improves the reconstructed images with a 42.92% and 67.12% lower root mean square error RMSE, 1.66% and 16.16% higher correlation coefficient CC, and 21.37% lower image error IE values in the simulation and experimental studies respectively as compared to the standard sensitivity matrix. The sectorial sensitivity matrix has an advantage of high performance even when the signal to noise ratio in an ECT system is low. Under such conditions, the present normalization method is incapable of calculating the normalized capacitance of the inclusion due to the low ratio of permittivity of the inclusion to that of the background. (paper)