Aiming at the ill posed and ill conditioned problems in image reconstruction when electrical capacitance tomography technology is applied to the detection of industrial industrial multiphase flow pipeline, an improved regularized half threshold algorithm is proposed. Taking L1 / 2 norm as the penalty function, the threshold operator in the semi threshold iterative algorithm used to solve L1 / 2 norm is improved. Taking the solution vector of Landweber algorithm with acceleration term as the correction vector, the improved semi threshold regularization model is introduced to optimize the accelerated Landweber algorithm. The experimental results show that the average correlation coefficient of the improved regularized half threshold algorithm in the reconstructed image is as high as 0. 91, the average image error is reduced to 0. 21, and the imaging speed is maintained at 0. 04 s. For complex flow pattern identification, the correlation coefficient of the improved algorithm is increased by 21. 67% and the relative error is reduced by 37. 01% compared with the traditional Landweber iterative algorithm; Compared with Tikhonov regularization algorithm, the correlation coefficient is increased by 22. 61% and the relative error is reduced by 37. 08% ; Compared with the half threshold algorithm, the correlation coefficient and relative error are increased by 14. 85% and reduced by 28. 26% respectively. The results show that the improved regularized half threshold algorithm has a good application prospect for ECT research.