Electrical capacitance tomography (ECT) is a potent image-based measurement technology for monitoring industrial processes, but low-quality images generally limit its application scope and measurement reliability. To increase the precision of reconstruction, in this study, a data-driven plug-and-play prior abstracted by a deep convolutional neural network (DCNN) and the sparseness prior of imaging objects, in form of regularizers, are jointly leveraged to generate a potent imaging model, in which the L₁ norm of the mismatch error acts as a data fidelity term (DFT) to weaken the sensitivity of estimation result to noisy input data. The DCNN is embedded into the split Bregman (SB) technique to generate a powerful computing scheme for solving the built imaging model and the fast iterative shrinkage-thresholding algorithm (FISTA) is applied to solve the sub-problems efficiently. Extensive numerical results verify that the proposed imaging technique has competitive reconstruction ability and better robustness in comparison with the state-of-the-art methods. This study demonstrates the validity and efficacy of the proposed algorithm in reducing reconstruction error. Most importantly, the research outcomes verify that the data-driven plug-and-play prior and the sparseness prior can be jointly embedded into the imaging model, leading to a remarkable decline in reconstruction error.

中文翻译：

---

具有先验的即插即用功能的电容层析成像计算成像方法

电容层析成像（ECT）是一种用于监视工业过程的基于图像的有效测量技术，但是低质量的图像通常会限制其应用范围和测量可靠性。为了提高重建的精度，在这项研究中，由深度卷积神经网络（DCNN）提取的数据驱动即插即用先验和成像对象的稀疏先验以正则化形式被联合利用以生成有效成像模型，其中L ₁失配误差的范数充当数据保真度项（DFT），从而削弱估计结果对嘈杂的输入数据的敏感性。将DCNN嵌入到分裂Bregman（SB）技术中，以生成强大的计算方案来解决构建的成像模型，并应用快速迭代收缩阈值算法（FISTA）有效地解决子问题。大量的数值结果证明，与最新方法相比，该成像技术具有竞争性的重建能力和更好的鲁棒性。这项研究证明了该算法在减少重构误差方面的有效性和有效性。最重要的是，研究结果验证了数据驱动的即插即用先验和稀疏先验可以共同嵌入到成像模型中，