摘要
电容层析成像系统在工业上用于管道内流体的实时监测，本文提出用神经网络的方法来代替传统的反投影算法进行图像重建，以解决传统算法存在的扭曲形变和空间精度受限的问题．文中用改进的ＢＰ算法训练，采用一个两层感知机网络，网络的输入是预处理过的测量电容矢量，输出直接对应到空间图像．实验结果表明使用这一方法。
Electrical capacitance tomography system is useful in real time imaging of multiphase flow in oil industry.The conventional back projection algorithm for capacitance tomography image reconstruction has the problems of distortion and limitation on spatial resolution. A neural method is proposed to replace the back projection method in this paper. We employ a two layer perceptron network and the training algorithm is the improved back propagation algorithm . The input vector of the network is the preprocessed capacitance measurements and the output corresponds to the spatial image. Experimental results illustrate that our method is practical and accurate for image reconstruction.
出处
《武汉大学学报（自然科学版）》
CSCD
1997年第1期107-112,共6页
Journal of Wuhan University(Natural Science Edition)
基金
国家自然科学基金
关键词
电容层析成像
神经网络
石油管道
在线监测
electrical capacitance tomography, neural network, perceptron, back propagation
