Research On Electrical Resistance Tomography Technology Based On Nonlinear Mapping

Electrical Resistance Tomography(ERT)is an emerging visual process detection technology with broad application prospects,which has the characteristics of non-invasive,safe,simple,and portable,cost-efficiency.However,due to its "soft field" effect,the reconstruction process is highly non-linear,ill-posed and ill-conditioned,resulting in serious degradation and low spatial resolution of the reconstructed image.Reducing the degradation of ERT reconstructed images and improving imaging accuracy have always been the focus and hot issues of ERT research at home and abroad.Aiming at the image degradation problem of ERT,this thesis focuses on the research of ERT reconstruction based on nonlinear mapping.Aiming at the problem of ERT image degradation,starting from the ERT image reconstruction principle and mathematical model,the mapping relationship between field distribution and boundary measurement voltage is analyzed.It is pointed out that in the ERT sensitive field,the linear approximation of the mapping relationship between the boundary voltage and the conductivity distribution is the main reason for the reconstructed image degradation,and it is also an important factor that causes the serious inconsistency of the sensitivity coefficient,which aggravates the image degradation and distortion.Aiming at the influence of ERT linearization solution on the reconstructed image degradation,the reconstruction error of using the uniform field sensitivity matrix is firstly analyzed,and an iterative dynamic reconstruction algorithm with nonlinear mapping relationship is proposed by alternately correcting the correlation between the reconstructed image and the sensitivity matrix.According to the correlation between the change of boundary voltage and the error introduced by the initial sensitivity matrix,the correction coefficient with penalty factor is obtained and used to modify the conductivity distributionand initial sensitivity matrix after previous reconstruction,and the modified sensitivity matrix is used for next image reconstruction.Experimental results show that the proposed algorithm can achieve a clearer boundary between the two media in the field with fewer artifacts and more accurate reconstructed distribution.For the difficulty of solving ERT nonlinear mapping by traditional algorithms,an ERT reconstruction algorithm based on Conditional Generative Adversarial Network(CGAN)is proposed.By learning the nonlinear mapping relationship between the boundary measurement and the target image,the proposed algorithm realizes the end-to-end application of deep learning from the measured value to the image,which could effectively improve the accuracy of the reconstructed image,and reveal more details of the reconstructed image,and achieve stronger practicability.According to the diversity of media distribution in ERT field,a data set generating method is proposed by randomizing medium distribution and the data set is established to realize effective learning of proposed algorithm.Experiments show that the data set generation method can simulate the diversity of media distribution well.The network is used to restore the degraded image,and the reconstructed image obtained is further optimized.Aiming at the problems of low sensitivity in the center of the ERT field and blurred reconstructed images,a center-electrode ERT sensor structure and ERT algorithm are proposed.The ERT sensor structure with a center electrode is used for excitation to improve the sensitivity of the field center.In order to overcome the imaging distortion near the electrode caused by the center electrode,a finite element-level image fusion reconstruction algorithm is proposed.The accuracy of the central field image is improved,and the effectiveness of the method is further verified by a manually generated data set.