为满足新型电力系统对能源协同输送走廊建设中介质分布的高精度监测需求，提出一种基于端到端神经网络的复值多频电容层析成像（complex-valued multi-frequency electrical capacitance tomography, CVMF-ECT）图像重建方法。首先，构建了一种新的背景预测网络，用于预测介质的背景分布，为初步成像提供先验约束。其次，利用图像重建强化网络对初步重建图像进行像素级分类，使每个像素对应一种介质类型，准确反映了管道中的介质分布及其特征，实现了高精度的图像重建。最后，采用结构相似度对成像质量进行量化评估，并与典型算法对比验证。仿真结果表明，经神经网络优化后的重建图像与目标图像间结构相似度均高于0.85，且在噪声干扰和频率变化的影响下仍能输出高精度的介质分布图像。该方法为高电导率下介质分布的高精度成像提供了一个新的技术思路。

In order to meet the demand for high-precision monitoring of medium distribution in the construction of collaborative energy transmission corridors for new power systems, a complex-valued multi-frequency electrical capacitance tomography (CVMF-ECT) based on an end-to-end neural network is proposed as an image reconstruction method. First, a new background prediction network is constructed for predicting the background distribution of the medium to provide priori constraints for preliminary imaging. Second, an image reconstruction enhancement network is used to classify the preliminary reconstructed images at the pixel level, so that each pixel corresponds to a media type, accurately reflecting the media distribution and its characteristics in the pipeline, and realizing high-precision image reconstruction. Finally, the structural similarity is used to quantitatively assess the imaging quality, and compared with typical algorithms for verification. The simulation results show that the structural similarity between the reconstructed image enhanced by neural network and the target image are all higher than 0.85, and the high-precision media distribution image can still be output under the influence of noise interference and frequency change. This method provides a new technical idea for high-precision imaging of medium distribution under high conductivity.