Abstract:Context. Multiconjugate adaptive optics (MCAO) can overcome atmospheric anisoplanatism to achieve high-resolution imaging with a large field of view (FOV). Atmospheric tomography is the key technology for MCAO. The commonly used modal tomography approach reconstructs the three-dimensional atmospheric turbulence wavefront aberration based on the wavefront sensor (WFS) detection information from multiple guide star (GS) directions. However, the atmospheric tomography problem is severely ill-posed. The incomplete GS coverage in the FOV coupled with the WFS detection error significantly affects the reconstruction accuracy of the three-dimensional atmospheric turbulence wavefront aberration, leading to a nonuniform aberration detection precision over the whole FOV. Aims. We propose an efficient approach for achieving accurate atmospheric tomography to overcome the limitations of the traditional modal tomography approach. Methods. We employed a deep-learning-based approach to the tomographic reconstruction of the three-dimensional atmospheric turbulence wavefront aberration. We propose an atmospheric tomography residual network (AT-ResNet) that is specifically designed for this task, which can directly generate wavefronts of multiple turbulence layers based on the Shack-Hartmann (SH) WFS detection images from multiple GS directions. The AT-ResNet was trained under different turbulence intensity conditions to improve its generalization ability. We verified the performance of the proposed approach under different conditions and compared it with the traditional modal tomography approach. Results. The well-trained AT-ResNet demonstrates a superior performance compared to the traditional modal tomography approach under different atmospheric turbulence intensities, various turbulence layer distributions, higher-order turbulence aberrations, detection noise, and reduced GSs conditions. The proposed approach effectively addresses the limitations of the modal tomography approach, leading to a notable improvement in the accuracy of atmospheric tomography. It achieves a highly uniform and high-precision wavefront reconstruction over the whole FOV. This study holds great significance for the development and application of the MCAO technology.

An ANN-based smart tomographic reconstructor in a dynamic environment

Deep tomography for the three-dimensional atmospheric turbulence wavefront aberration

Comparative Study of Neural Network Frameworks for the Next Generation of Adaptive Optics Systems

Dynamic-excitation-based Steady-State Non-Line-of-sight Imaging Via Multi-Branch Convolutional Neural Network

Singular Value and Frame Decomposition-based Reconstruction for Atmospheric Tomography

Efficient reconstruction method for ground layer adaptive optics with mixed natural and laser guide stars

Enhancing Dynamic CT Image Reconstruction with Neural Fields and Optical Flow

Making the unmodulated Pyramid wavefront sensor smart. Closed-loop demonstration of neural network wavefront reconstruction with MagAO-X

Neural Fields for Adaptive Photoacoustic Computed Tomography

Solution reconstruction for computational fluid dynamics via artificial neural network

Physics to the Rescue: Deep Non-line-of-sight Reconstruction for High-speed Imaging

Deep-Learning-Based Approach in Imaging Radiometry by Aperture Synthesis: An Alias-Free Method

3D Data Processing and Entropy Reduction for Reconstruction from Low-Resolution Spatial Coordinate Clouds in a Technical Vision System

A Memory-Efficient Dynamic Image Reconstruction Method using Neural Fields

NeDF: neural deflection fields for sparse-view tomographic background oriented Schlieren

Non-line-of-sight imaging with adaptive artifact cancellation

When linear inversion fails: neural-network optimization for sparse-ray travel-time tomography of a volcanic edifice

Enhancing Open-Loop Wavefront Prediction in Adaptive Optics through 2D-LSTM Neural Network Implementation

Deep-Learning Image Reconstruction for Real-Time Photoacoustic System

Sparse-view Signal-domain Photoacoustic Tomography Reconstruction Method Based on Neural Representation