摘要
描述了 BP网络应用于 EMT图像重建的基本原理和简易模型 .网络的输入是传感器二维磁场正问题的有限元仿真得到的测量数据 ,网络的输出是物体空间各剖分单元的 0 - 1状态值 ,网络用共轭梯度法改进的误差逆传播算法进行训练 .简单物流的图像重建结果表明 ,网络模型在原理上是可行的 。
The paper presents a BP neural network approach applied to image rconstruction of electromagnetic tomography (EMT).The input of the network was measurements of the EMT sensor,which were obtained by solving the forward electroomagnetic problem of the sensor by finite element simulation.State values (0 1) of all the elements in the object space were taken as the output of the network.The network was trained with the error back propagation algorithm improved by such means as the conjugate gradient optimization method.The initial results of image reconstruction for simple and easy two component flows show that the model feasible is in principle,which gives a basis to further analysis and research.
出处
《天津大学学报》
EI
CAS
CSCD
2000年第2期138-143,共6页
Journal of Tianjin University(Science and Technology)
基金
国家自然科学基金资助项目! (695740 2 1和 598770 1 8)
关键词
过程成像
电磁场
BP网络
图像重建
电磁层析成像
process tomography
electromagnetic fields
finite element methods
BP networks
image reconstruction
sensor techniques
two phase flow measurements
