Abstract
We present a quantitative study of coherent array imaging of remote sources in randomly perturbed waveguides with bounded cross-section. We study how long range cumulative scattering by perturbations of the boundary and the medium impedes the imaging process. We show that boundary scattering effects can be mitigated with filters that enhance the coherent part of the data. The filters are obtained by optimizing a measure of quality of the image. The point is that there is an optimal trade-off between the robustness and resolution of images in such waveguides, which can be found adaptively, as the data are processed to form the image. Long range scattering by perturbations of the medium is harder to mitigate than scattering by randomly perturbed boundaries. Coherent imaging methods do not work and more complex incoherent methods, based on transport models of energy, should be used instead. Such methods are neither useful, nor needed in waveguides with perturbed boundaries. We explain all these facts using rigorous asymptotic stochastic analysis of the wave field in randomly perturbed waveguides. We also analyze the adaptive coherent imaging method and obtain a quantitative agreement with the results of numerical simulations.
Introduction
Imaging a remote source within a waveguide is a fundamental problem in fields ranging from underwater acoustics to fiber optic communications. In an ideal waveguide with perfectly smooth boundaries and a homogeneous medium, waves propagate in predictable patterns known as modes. However, real-world environments are rarely ideal. Small-scale fluctuations in the boundary geometry or the refractive index of the medium lead to cumulative scattering over long distances. This scattering gradually degrades the coherent signal—the part of the wave field that carries direct information about the source location—and replaces it with incoherent noise or reverberations.
This study addresses two critical questions: How do different types of randomness (boundary vs. medium) affect the decay of coherence? And can we adaptively process array data to recover source information even in the presence of significant scattering?
Methodology and System Model
The researchers employ a combination of asymptotic stochastic analysis and high-fidelity numerical simulations. The wave field is modeled as a superposition of waveguide modes. In a random waveguide, these modes "couple," meaning energy is transferred between them as they scatter off imperfections.
Adaptive Coherent Imaging
The core innovation is an adaptive filtering approach. Standard imaging often treats all received data equally, but in a random waveguide, some modes retain coherence longer than others. The authors introduce a "figure of merit"—defined as the ratio of the image's peak intensity to its total energy—to quantify image quality. By using an optimization routine (specifically the fmincon function in MATLAB), they find the optimal weights for each mode to maximize this ratio. This process effectively filters out the "noisy" modes while preserving those that still contain coherent source information.
Numerical Validation
To validate their theoretical models, the authors used the "Montjoie" finite element code to simulate 2D waveguides. They generated synthetic datasets for two primary scenarios:
- Boundary Perturbations: Modeled as a Matérn-7/2 stationary random process affecting the waveguide walls.
- Medium Perturbations: Modeled using Gaussian wave speed fluctuations within the waveguide interior. The imaging array consisted of 39 receivers placed at distances ranging from $25\lambda_o$ to $150\lambda_o$ from the source.
Key Insights: Boundary vs. Medium Randomness
The study reveals a fundamental physical difference between scattering caused by boundary roughness and scattering caused by medium fluctuations.
The Boundary Divide
In waveguides with random boundaries, "fast" modes (those with low mode numbers that travel nearly parallel to the waveguide axis) interact with the walls less frequently. Consequently, they remain coherent over much longer distances than "slow" (high-order) modes, which bounce frequently off the perturbed boundaries. This disparity allows the adaptive filter to be highly effective: by suppressing the high-order modes that have turned into noise, the system can maintain a clear, albeit lower-resolution, image of the source.
The Medium Challenge
In contrast, random fluctuations within the medium affect all modes more uniformly. The "scattering mean free path"—the distance over which a mode loses its coherence—is much shorter for the fast modes in the medium case compared to the boundary case. Because all modes lose coherence at roughly the same rate, there is no "clean" subset of data for the adaptive filter to isolate. Coherent imaging fails entirely once the source distance exceeds approximately $50\lambda_o$.
Results and Discussion
The results demonstrate a clear trade-off between image resolution and robustness. High-resolution imaging requires the inclusion of many modes, but high-order modes are the most susceptible to scattering. The adaptive weighting method successfully finds the "sweet spot" for boundary-perturbed waveguides, significantly outperforming standard imaging at ranges up to $100\lambda_o$.
Furthermore, the study shows that in random media, coherent imaging fails long before the energy reaches an "equipartition" state (where energy is evenly distributed among modes). For random boundaries, however, the failure of coherent imaging coincides closely with the point where the wave field loses all memory of its initial state.
Conclusion
This research provides a rigorous framework for understanding the limits of source imaging in complex environments. While adaptive coherent methods are powerful tools for mitigating boundary-induced noise, they are insufficient for waveguides with significant medium fluctuations. In the latter case, the authors suggest that more complex incoherent methods based on energy transport models are required. Future work may involve extending these 2D findings to 3D bounded waveguides and refining the "figure of merit" with prior information about source characteristics.
Summary updated