摘要
为解决电容层析成像中软场效应导致重建图像精度低的问题,提出了一种基于并联自适应残差网络与卷积注意力机制的图像重建算法。通过引入并联自适应残差模块提取丰富的特征层信息,再利用压缩激励网络调整各通道的权重系数,达到过滤冗余信息的效果,引入卷积注意力机制学习浅层特征的通道和空间信息,将卷积注意力机制通道与并联自适应残差网络进行特征融合以补偿损失的浅层特征和空间信息。仿真结果表明,相比LBP算法、Landweber迭代算法、1D CNN算法,改进算法有效提高了重建质量。
In order to solve the problem that soft effect in capacitance tomography should lead to low accuracy of reconstructed image,a parallel adaptive residual network with convolutional attention mechanism image reconstruction algorithm is proposed.The parallel adaptive residual module is used to extract the rich feature layer information.The weight coefficients of each channel are adjusted by the squeeze and excitation networks to filter the redundant information.The convolutional attention mechanism is used to learn the channel and space information of the shallow feature.The feature fusion between the convolutional attention mechanism channel and the parallel adaptive residual network is proposed to compensate the lost shallow features and spatial information.Simulation results show that compared with LBP algorithm,Landweber iterative algorithm and 1DCNN algorithm,the improved algorithm can guarantee real-time performance and improve reconstruction quality effectively.
作者
马敏
吴环
MA Min;WU Huan(College of Electronic Information and Automation,Civil Aviation University of China,Tianjin 300300,China)
出处
《计量学报》
CSCD
北大核心
2024年第2期214-221,共8页
Acta Metrologica Sinica
基金
国家自然科学基金(61871379)。
关键词
多相流测量
电容层析成像
图像重建
并联自适应残差网络
卷积注意力机制
multiphase flow measurement
electrical capacitance tomography
image reconstruction
parallel adaptive residual network
convolution block attention module
