计量学报2024，Vol.45Issue(8)：1139-1146,8.DOI:10.3969/j.issn.1000-1158.2024.08.08
基于多尺度自适应网络的ECT图像重建
Image Reconstruction of ECT Based on Multi-scale Adaptive Network
摘要
Abstract
In order to solve the nonlinear ill-posed inverse problem of electrical capacitance tomography(ECT),a multi-scale adaptation network(MSANet)model is proposed,which achieves the fusion of multi-scale features in a more fine-grained dimension and has a relatively small number of model parameters.By constructing a tree structure within a single residual block to form a multi-scale feature fusion module,MSANet achieves more robustness and lower computational parameters.Furthermore,by adding an adaptive spatial threshold module,the reconstruction accuracy of the images is further improved.Compared with linear back projection(LBP)algorithm,Landweber iterative algorithm,and commonly used deep learning methods,this method has the smallest average relative error and the largest average correlation coefficient,with 0.181 and 0.967,respectively.关键词
多相流测量/机器视觉/电容层析成像/图像重建/深度学习/多尺度/自适应网络Key words
multiphase flow measurement/machine vision/electrical capacitance tomography/image reconstruction/deep learning/multi-scale/adaptive network分类
通用工业技术引用本文复制引用
张立峰,常恩健..基于多尺度自适应网络的ECT图像重建[J].计量学报,2024,45(8):1139-1146,8.基金项目
国家自然科学基金(61973115) （61973115）
