基于近似L0范数的电容层析成像敏感场优化算法  

Electrical Capacitance Tomography Sensitivity Field Optimization Algorithm Based on Approximate L Norm

作　　者：马敏 刘一斐 王世喜 Ma Min;Liu Yifei;Wang Shixi(College of Electronic Information and Automation,Civil Aviation University of China,Tianjin300300,China)

机构地区：[1]中国民航大学电子信息与自动化学院,天津300300

出　　处：《激光与光电子学进展》2021年第12期264-273,共10页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(61871379)。

摘　　要：针对电容层析成像(ECT)逆问题求解过程中欠定性的问题,引入一种近似L0范数的稀疏正则化算法以获得稀疏解向量。针对敏感场灵敏度分布不均所引起的成像质量问题,提出一种可迭代敏感场灵敏度梯度优化方法,该方法以敏感场各有限元为核心将敏感场划分为若干个区域,提取围绕该有限元区域内的敏感度数据并进行均值滤波,所得数值返回该有限元中并作为下一滤波区域的参数,通过循环滤波可逐渐降低敏感场中心区域与边缘区域的灵敏度梯度。将优化后的灵敏度梯度优化方法与近似L0算法结合以验证所提算法的可行性。结果表明,与传统Landweber算法相比,所提算法将重建图像的相对误差降低至0.24,相关系数提升至0.91,实际的静态实验也证明该算法的有效性。Aiming at the under-qualitative problem in solving the inverse problem of electrical capacitance tomography,a sparse regularization algorithm that approximates the L0 norm is introduced to obtain the sparse solution vector.An iterable sensitivity gradient optimization method of the sensitive field is proposed to address the imaging quality problem caused by the uneven sensitivity distribution of the sensitive field.This method uses the finite elements of the sensitive field as the core to divide the sensitive field into several regions and the data of sensitivity in the region around the core finite element is extracted for mean filtering.And the filtered data is returned to the core finite elements and used as the parameters in the next filtering area.Cyclic filtering can gradually reduce the sensitivity gradient between the center area and edge area of the sensitive field.The sensitivity gradient optimization method is combined with the approximate L0 algorithm to verify the feasibility of the proposed algorithm.The results show that compared with the traditional Landweber algorithm,the proposed algorithm reduces the relative error of a reconstructed image to 0.24 and the correlation coefficient to 0.91.The actual static experiment also proves the effectiveness of the proposed algorithm.

关 键 词：图像处理 电容层析成像 L0范数 稀疏优化 敏感场 灵敏度优化 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]