Skip to main content
QUICK REVIEW

[论文解读] Phased and phaseless domain reconstruction in inverse scattering problem via scattering coefficients

Habib Ammari, Yat Tin Chow|arXiv (Cornell University)|Oct 14, 2015
Microwave Imaging and Scattering Analysis参考文献 35被引用 5
一句话总结

本文针对基于散射系数的无相位逆散射问题,提出了稳定性分析与重建算法,对比了有相位与无相位重建方法。提出了一种基于条件数的策略以优化无相位测量,表明无相位重建虽显著更病态,但通过适当的采样与正则化可实现鲁棒结果,数值实验验证了在5%噪声下L²误差低于7%。

ABSTRACT

In this work we shall review the (phased) inverse scattering problem and then pursue the phaseless reconstruction from far-field data with the help of the concept of scattering coefficients. We perform sensitivity, resolution and stability analysis of both phased and phaseless problems and compare the degree of ill-posedness of the phased and phaseless reconstructions. The phaseless reconstruction is highly nonlinear and much more severely ill-posed. Algorithms are provided to solve both the phased and phaseless reconstructions in the linearized case. Stability is studied by estimating the condition number of the inversion process for both the phased and phaseless cases. An optimal strategy is suggested to attain the infimum of the condition numbers of the phaseless reconstruction, which may provide an important guidance for efficient phaseless measurements in practical applications. To the best of our knowledge, the stability analysis in terms of condition numbers are new for the phased and phaseless inverse scattering problems, and are very important to help us understand the degree of ill-posedness of these inverse problems. Numerical experiments are provided to illustrate the theoretical asymptotic behavior, as well as the effectiveness and robustness of the phaseless reconstruction algorithm.

研究动机与目标

  • 分析基于散射系数的有相位与无相位逆散射重建的敏感性、分辨率与稳定性。
  • 通过比较有相位与无相位问题的条件数,量化其病态程度。
  • 在线性化条件下,为有相位与无相位情况开发高效的重建算法。
  • 提出一种最优测量策略,以最小化无相位重建的条件数,为实际数据采集提供指导。

提出的方法

  • 通过对比函数 q(x) = ε*χ_D(x) 建模声学介质散射,基于Helmholtz方程并利用远场测量。
  • 从远场散射模式中推导出散射系数,用于线性化有相位与无相位两种情况下的逆问题。
  • 通过估计反演过程的条件数来评估病态性,结果表明无相位重建显著更病态。
  • 采用Tikhonov型正则化,结合L¹与L²惩罚项,以稳定无相位重建。
  • 通过域扰动的傅里叶模态展开实现数值算法,用于重建形状与对比度。
  • 通过最小化条件数推导出最优测量配置,为无相位数据采集提供实用指导。

实验结果

研究问题

  • RQ1无相位重建问题的条件数与有相位问题相比如何?这对其相对病态性有何含义?
  • RQ2能否通过散射系数与适当的正则化有效稳定无相位重建问题?
  • RQ3为最小化重建误差,无相位测量所需的最优波数数量与分布是什么?
  • RQ4不同的正则化策略(L¹与L²)如何影响无相位重建的精度与鲁棒性?
  • RQ5在噪声条件下,仅靠无相位数据在多大程度上可重建包含物的形状与对比度?

主要发现

  • 无相位重建比有相位重建显著更病态,表现为条件数高得多。
  • 所提出的最优测量策略有效最小化了无相位重建的条件数,为高效数据采集提供了实用框架。
  • 数值实验表明,在最优采样与正则化下,无相位重建在5%噪声下相对L²误差可低至0.57%。
  • 当采用次优测量集时,重建质量下降,最坏情况下误差上升至14.84%。
  • L¹正则化在特定情况下表现优异,即使精确解并非稀疏,也显示出鲁棒性。
  • 该方法成功以高保真度重建了复杂花形形状,证明了其在严重病态条件下的有效性。

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。