基于K-SVD字典学习的电容层析成像三维图像重建算法  

Three Dimensional Image Reconstruction Based on K-SVD for Electrical Capacitance Tomography

作　　者：王韵然 陈德运[1] 王莉莉[1] WANG Yun-ran;CHEN De-yun;WANG Li-li(School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080, China)

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《哈尔滨理工大学学报》2020年第4期95-100,共6页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(60572153,60972127);高等学校博士学科点专项科研基金(200802140001);黑龙江省自然科学基金(QC2012C059);哈尔滨市科技创新人才研究专项资金(2014RFXXJ022);黑龙江省教育厅计划项目(11541040,12531094).

摘　　要：电容层析成像(ECT,electrical capacitance tomography)是两相流参数测量的有效方法之一,电容层析成像三维图像重建是基于多层电极结构的传感器系统,通过测量同层及异层极板之间的电容值,建立新的冗余灵敏度矩阵来映射介电常数分布和传感器读数之间的关系。作为高度非线性的简单线性化问题,存在被成像的物体的位置和尺寸不匹配。为了克服这些问题,采用贪婪稀疏化的方法对三维图像进行重建,利用K-SVD字典学习方法和非线性仿真构造冗余灵敏度矩阵,通过对冗余灵敏度矩阵真实数据的训练,能够更好地捕获被测物体的典型结构。通过仿真实验比较,所提出的方法也可以改善重建的图像质量。ECT(Electrical Capacitance Tomography)is one of the effective methods to measure the parameters of two-phase flow.The three-dimensional image reconstruction of capacitance tomography is a sensor system based on multi-layer electrode structure.A new redundant sensitivity matrix is established to map the relationship between the dielectric constant distribution and the sensor reading by measuring the capacitance values between the same layer and the heterogeneous plate.As a highly nonlinear simple linearization problem,there are mismatches in the position and size of the object being imaged.To overcome these problems,and the greedy sparsity method is used to reconstruct the 3d image.The redundant sensitivity matrix is constructed by using k-svd dictionary learning method and nonlinear simulation.The typical structure of the object under test can be better captured by training the real data of the redundant sensitivity matrix.The proposed method can also improve the image quality of the reconstruction by comparing the simulation experiments.

关 键 词：三维ECT传感器 灵敏度矩阵 三维图像重建 字典学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]