EIT Imaging Method Based on B-Spline Feature Extraction and Sparse Reconstruction
Abstract
Electrical Impedance Tomography (EIT) offers several advantages, including high temporal resolution, radiation-free operation, and low cost. However, its inverse problem is severely ill-posed, leading to low imaging accuracy and strong artifacts. These drawbacks limit its practical applications. To address this issue, this article proposes a novel EIT image reconstruction method based on B-spline feature extraction and sparse reconstruction. First, effective features from the conductivity distribution are extracted by B-spline kernels. Observation data are augmented by this process, and the ill-posedness of the inverse problem is mitigated accordingly. Subsequently, an enhanced composite observation model is constructed, and the image reconstruction task is transformed into an L0-norm optimization problem. This problem is solved by the adaptive spatial constrained orthogonal matching pursuit (ASCOMP) algorithm. ASCOMP can automatically determine the number of iterations, enhance spatial continuity, and produce globally optimal coefficients within each selected support set using a convex optimization subproblem. Finally, experimental validation was conducted using a 16-electrode hardware system. In simulations of six typical flow patterns, the new method achieved a correlation coefficient (CC) as high as 0.9272 and a relative error (RE) as low as 0.2839, with average values of 0.8933 and 0.3752, respectively, outperforming all other algorithms. Hardware experiments also demonstrated that it effectively suppresses artifacts and preserves target boundaries, performing particularly well in complex multitarget scenarios. In summary, this method provides an effective solution to the EIT inverse problem.