Deep optics framework utilizing a liquid crystal microlens array for dynamic wavefront sensing and correction in atmospheric turbulence

Anisotropic atmospheric turbulence, which distorts wavefronts, poses challenges to advancements in fields such as astronomy and remote sensing. Existing reconstruction methods often fail to adequately capture its spatiotemporal variability, leading to suboptimal performance. To address this, we propose a spatiotemporal quantization-aware deep optics framework. This framework integrates light field physics models with a data-driven neural network via a liquid crystal microlens array (LC-MLA), offering a solution to dynamic turbulence issues. The framework constructs a spatiotemporal transfer function that integrates wavefront sensing, high-dimensional spatiotemporal quantization-aware turbulence modeling, and neural network characterization. By synergizing dynamic light field data with electronically tunable LC-MLA, our approach optimizes hardware and software parameters for efficient, precise correction of multi-field, multi-order aberrations. In experiments, this approach optimizes real-time turbulence correction and offers a 17.5% improvement in peak signal-to-noise ratio compared to previous methods.