We present a novel machine-learning approach for detecting faint point
sources in high-contrast adaptive optics imaging datasets. The most widely used
algorithms for primary subtraction aim to decouple bright stellar speckle noise
from planetary signatures by subtracting an approximation of the temporally
evolving stellar noise from each frame in an imaging sequence. Our approach
aims to improve the stellar noise approximation and increase the planet
detection sensitivity by leveraging deep learning in a novel direct imaging
post-processing algorithm. We show that a convolutional autoencoder neural
network, trained on an extensive reference library of real imaging sequences,
accurately reconstructs the stellar speckle noise at the location of a
potential planet signal. This tool is used in a post-processing algorithm we
call Direct Exoplanet Detection with Convolutional Image Reconstruction, or
ConStruct. The reliability and sensitivity of ConStruct are assessed using real
Keck/NIRC2 angular differential imaging datasets. Of the 30 unique point
sources we examine, ConStruct yields a higher S/N than traditional PCA-based
processing for 67$\%$ of the cases and improves the relative contrast by up to
a factor of 2.6. This work demonstrates the value and potential of deep
learning to take advantage of a diverse reference library of point spread
function realizations to improve direct imaging post-processing. ConStruct and
its future improvements may be particularly useful as tools for post-processing
high-contrast images from the James Webb Space Telescope and extreme adaptive
optics instruments, both for the current generation and those being designed
for the upcoming 30 meter-class telescopes.

我们提出了一种新颖的机器学习方法，用于检测高对比度自适应光学成像数据中的微弱点源。通过利用深度学习在一种新颖的直接成像后处理算法中，我们的方法旨在改进恒星噪声的近似并提高行星探测灵敏度。我们展示了一个卷积自编码神经网络在广泛的实际成像序列参考库上进行训练，能够准确重构潜在行星信号位置的恒星斑点噪声。我们称这个工具为 ConStruct，通过使用真实的 Keck/NIRC2 角差分成像数据集评估了 ConStruct 的可靠性和灵敏度。在我们考察的 30 个独特的点源中，ConStruct 对 67％的情况产生比传统基于 PCA 的处理更高的信噪比，并且相对对比度提高了最多 2.6 倍。这项工作展示了深度学习利用多样的点扩散函数实现来改进直接成像后处理的价值和潜力。ConStruct 及其未来的改进可能特别适用于处理来自詹姆斯・韦伯太空望远镜和极端自适应光学仪器的高对比度图像，无论是对于当前的望远镜还是设计中的 30 米级望远镜。

使用深度卷积图像重建的直接系外行星检测（ConStruct）：一种用于后处理高对比度图像的新算法

Direct Exoplanet Detection Using Deep Convolutional Image Reconstruction  (ConStruct): A New Algorithm for Post-Processing High-Contrast Images

We present the measurement of the Hubble Constant, $H_0$, with three strong
gravitational lens systems. We describe a blind analysis of both PG1115+080 and
HE0435-1223 as well as an extension of our previous analysis of RXJ1131-1231.
For each lens, we combine new adaptive optics (AO) imaging from the Keck
Telescope, obtained as part of the SHARP AO effort, with Hubble Space Telescope
(HST) imaging, velocity dispersion measurements, and a description of the
line-of-sight mass distribution to build an accurate and precise lens mass
model. This mass model is then combined with the COSMOGRAIL measured time
delays in these systems to determine $H_{0}$. We do both an AO-only and an
AO+HST analysis of the systems and find that AO and HST results are consistent.
After unblinding, the AO-only analysis gives $H_{0}=82.8^{+9.4}_{-8.3}~\rm
km\,s^{-1}\,Mpc^{-1}$ for PG1115+080, $H_{0}=70.1^{+5.3}_{-4.5}~\rm
km\,s^{-1}\,Mpc^{-1}$ for HE0435-1223, and $H_{0}=77.0^{+4.0}_{-4.6}~\rm
km\,s^{-1}\,Mpc^{-1}$ for RXJ1131-1231. The joint AO-only result for the three
lenses is $H_{0}=75.6^{+3.2}_{-3.3}~\rm km\,s^{-1}\,Mpc^{-1}$. The joint result
of the AO+HST analysis for the three lenses is $H_{0}=76.8^{+2.6}_{-2.6}~\rm
km\,s^{-1}\,Mpc^{-1}$. All of the above results assume a flat $\Lambda$ cold
dark matter cosmology with a uniform prior on $\Omega_{\textrm{m}}$ in [0.05,
0.5] and $H_{0}$ in [0, 150] $\rm km\,s^{-1}\,Mpc^{-1}$. This work is a
collaboration of the SHARP and H0LiCOW teams, and shows that AO data can be
used as the high-resolution imaging component in lens-based measurements of
$H_0$. The full time-delay cosmography results from a total of six strongly
lensed systems are presented in a companion paper.

利用强引力透镜系统的时间延迟宇宙学结果，结合自适应光学成像、质量模型和测量并构建了哈勃常数 $H_{0}$ 的高精度和准确模型，并发现自适应光学数据可以用作镜像测量中的高分辨率成像组件。