基于BP神经网络的ECT图像重建算法  

Image Reconstruction Algorithm for Electrical Capacitance Tomography Based on BP Neural Network

作　　者：马敏[1] 高振福[1] 王化祥[2] 

机构地区：[1]中国民航大学航空自动化学院,天津300300 [2]天津大学电气自动化工程学院,天津300072

出　　处：《计量学报》2013年第6期524-528,共5页Acta Metrologica Sinica

基　　金：国家自然科学基金青年基金（61102096,61201350）;天津市自然科学基金（11JCYBJC06900）.

摘　　要：针对电容层析成像图像重建问题的病态性，利用COMSOL软件建立系统模型，并结合MATLAB实现正问题的求解。依据BP神经网络所具有的理想的非线性映射和联想记忆功能实现了由检测电容值到重建图像灰度值之间的非线性映射，避免了传统算法中对灵敏度矩阵求解的繁琐，克服了因线性化处理所导致的成像精度低的缺点。在MATLAB平台下，采用2种滤波方法进行滤波，对图像增强修复，提高了图像质量。Considering the electrical capacitance tomography image reconstruction is an ill-posed problem, the system model is set up in COMSOL software, resolving the Forward problem. BP neural network is applied to realize the nonlinear mapping between the capacitance values and the reconstructed image grey values based on the nonlinear mapping and the association memory ability. Compared with the traditional ones, the algorithm avoids the trivial solution for the sensitivity matrix, and the linearization that reduces the imaging accuracy. Two filtering methods are proposed to improve image quality, and the simulation was completed in the MATLAB workbench.

关 键 词：计量学 电容层析成像 BP神经网络 图像重建 滤波 仿真 

分 类 号：B937[哲学宗教—宗教学]