基于评价指标的EIT算法参数选择方法研究

摘要

为客观准确地选择EIT算法参数，采用8种评价参数指标对共轭梯度算法的迭代次数和Tikhonov正则化算法的正则化因子的选择进行研究．首先，构建 EIT 正问题模型，并求得两种算法的逆问题解；然后，根据8种评价参数的定义，获得图像重建时的最佳参考值范围：共轭梯度迭代算法的迭代次数在70～80次，正则化因子的取值范围为0.01～0.1．为了验证上述结论，重建 EIT 图像，并进行对比分析．结果表明：基于评价指标获得参数重建图像的效果更令人满意；针对参数选择，提出的方法可为其他电阻抗图像重建算法提供一种客观的评价依据，为 EIT 图像质量评价体系的构建奠定基础．%To select the parameters of EIT algorithm objectively and accurately,a novel approach wasproposed to decide the number of iterations of the conjugate gradient(CG)algorithm and the regularization factors of Tikhonov regulariza-tion(TR)algorithm.Eight evaluation criterion were employed to evaluate the performances.Firstly,we encoded CG and TR algorithms respectively based on a circle EIT forward model,and obtained the solutions of each algorithm.Secondly, according to the definition of the eight evaluation criterion,the optimal reference values of the algorithms were obtained.The iteration number of the CG algorithm is 70-80 times,and the regularization factor of TR is suggested between 0.01-0.1.Thirdly,the EIT images were reconstructed with and without the reference value and compared.The quality of the im-age with the reference value is more satisfactory.After modification,the proposed method provides an objective assessment for the parameter selection of the electrical impedance image reconstruction algorithm,which lays a foundation for EIT im-age quality evaluation system.