A team of astronomers has employed a cutting-edge, artificial intelligence-assisted technique to uncover rare astronomical phenomena within archived data from NASA’s Hubble Space Telescope . The team analyzed nearly 100 million image cutouts from the Hubble Legacy Archive, each measuring just a few dozen pixels (7 to 8 arcseconds) on a side. They identified more than 1,300 objects with an odd appearance in just two and a half days — more than 800 of which had never been documented in scientific literature.
一个天文学家团队采用了一种尖端的人工智能辅助技术，在美国宇航局哈勃太空望远镜的存档数据中发现了罕见的天文现象。该团队分析了哈勃遗产档案中的近 1 亿张图像剪切图，每张图像的一侧仅测量几十个像素（7 到 8 角秒）。他们在短短两天半内识别出了 1,300 多个外观奇怪的物体，其中 800 多个物体从未在科学文献中记录过。
Most of the anomalies were galaxies undergoing mergers or interactions, which exhibit unusual morphologies or trailing, elongated streams of stars and gas. Others were gravitational lenses, where the gravity of a foreground galaxy distorts spacetime and bends light from a background galaxy into arcs or rings. Additional discoveries included galaxies with massive star-forming clumps, jellyfish-looking galaxies with gaseous “tentacles,” and edge-on planet-forming disks in our own galaxy resembling hamburgers. Remarkably, several dozen objects defied existing classification schemes entirely.
大多数异常现象是正在合并或相互作用的星系，它们表现出不寻常的形态或拖尾、拉长的恒星和气体流。其他的是引力透镜，其中前景星系的引力扭曲时空并将来自背景星系的光弯曲成弧形或环形。其他发现包括具有巨大恒星形成团块的星系、具有气态“触手”的水母状星系，以及我们银河系中类似汉堡包的边缘行星形成盘。值得注意的是，有几十个物体完全违背了现有的分类方案。
Identifying such a diverse array of rare objects within the vast and growing repository of Hubble and other telescope data presents a formidable challenge. Never in the history of astronomy has such a volume of observational data been available for analysis.
在庞大且不断增长的哈勃和其他望远镜数据存储库中识别如此多样化的稀有物体是一项艰巨的挑战。在天文学史上，从未有过如此大量的观测数据可供分析。
To address this challenge, researchers David O’Ryan and Pablo Gómez of ESA (the European Space Agency) developed an AI tool capable of inspecting millions of astronomical images in a fraction of the time required by human experts. Their neural network, named AnomalyMatch, was trained to detect rare and unusual objects by recognizing patterns in data — mimicking the way the human brain processes visual information.
为了应对这一挑战，欧洲航天局 (ESA) 的研究人员 David O’Ryan 和 Pablo Gómez 开发了一种人工智能工具，能够在人类专家所需时间的一小部分内检查数百万张天文图像。他们的神经网络名为 AnomalyMatch，经过训练可以通过识别数据模式来检测稀有和不寻常的物体——模仿人脑处理视觉信息的方式。
“Archival observations from the Hubble Space Telescope now span 35 years, offering a rich dataset in which astrophysical anomalies may be hidden,” said David O’Ryan, lead author of the study published in Astronomy & Astrophysics .
“哈勃太空望远镜的档案观测现已跨越 35 年，提供了丰富的数据集，其中可能隐藏着天体物理异常，”发表在《天文学与天体物理学》上的这项研究的主要作者 David O’Ryan 说。
Traditionally, anomalous images are discovered through manual inspection or serendipitous observation. While expert astronomers excel at identifying unusual features, the sheer volume of Hubble data makes comprehensive manual review impractical. Citizen science initiatives have helped expand the scope of data analysis, but even these efforts fall short when faced with archives as extensive as Hubble’s or those from wide-field survey telescopes like Euclid, an ESA mission with NASA contributions.
传统上，异常图像是通过手动检查或偶然观察发现的。虽然天文学家擅长识别不寻常的特征，但哈勃数据的庞大数量使得全面的手动审查变得不切实际。公民科学倡议帮助扩大了数据分析的范围，但即使是这些努力，在面对像哈勃望远镜或像欧几里得这样的宽视场巡天望远镜（欧几里德是欧空局的一个任务，美国宇航局也有贡献）这样广泛的档案时，也显得不够。
The work by O’Ryan and Gómez represents a significant advancement. By applying AnomalyMatch to the Hubble Legacy Archive, they conducted the first systematic search for astrophysical anomalies across the entire dataset. After the algorithm flagged likely candidates, the researchers manually reviewed the top-rated sources and confirmed more than 1,300 as true anomalies.
O’Ryan 和 Gómez 的工作代表了重大进步。通过将 AnomalyMatch 应用于哈勃遗产档案，他们对整个数据集中的天体物理异常进行了首次系统搜索。在算法标记出可能的候选者后，研究人员手动审查了最受好评的来源，并确认 1,300 多个为真正的异常。
“This is a powerful demonstration of how AI can enhance the scientific return of archival datasets,” Gómez said. “The discovery of so many previously undocumented anomalies in Hubble data underscores the tool’s potential for future surveys.”
“这是人工智能如何增强档案数据集的科学回报的有力证明，”戈麦斯说。 “在哈勃数据中发现了如此多以前未记录的异常现象，凸显了该工具在未来调查中的潜力。”
Hubble is just one of many astronomical archives poised to benefit from AI-driven analysis. Facilities such as NASA’s upcoming Nancy Grace Roman Space Telescope , a well as ESA’s Euclid and the National Science Foundation and Department of Energy’s Vera C. Rubin Observatory , will generate unprecedented volumes of data. Tools like AnomalyMatch will be essential for navigating this data deluge, enabling astronomers to uncover new and unexpected phenomena — and perhaps even objects never before seen in the universe.
哈勃只是众多有望受益于人工智能驱动分析的天文档案之一。美国宇航局即将推出的南希·格雷斯·罗马太空望远镜、欧洲航天局的欧几里德、国家科学基金会和能源部的维拉·C·鲁宾天文台等设施将产生前所未有的数据量。像 AnomalyMatch 这样的工具对于驾驭这场数据洪流至关重要，使天文学家能够发现新的、意想不到的现象——甚至可能是宇宙中从未见过的物体。
The Hubble Space Telescope has been operating for over three decades and continues to make ground-breaking discoveries that shape our fundamental understanding of the universe. Hubble is a project of international cooperation between NASA and ESA. NASA’s Goddard Space Flight Center in Greenbelt, Maryland, manages the telescope and mission operations. Lockheed Martin Space, based in Denver, also supports mission operations at Goddard. The Space Telescope Science Institute in Baltimore, which is operated by the Association of Universities for Research in Astronomy, conducts Hubble science operations for NASA.
哈勃太空望远镜已经运行了三十多年，并不断做出突破性的发现，塑造我们对宇宙的基本理解。哈勃是 NASA 和 ESA 之间的国际合作项目。美国宇航局位于马里兰州格林贝尔特的戈达德太空飞行中心负责管理望远镜和任务操作。总部位于丹佛的洛克希德·马丁航天公司也支持戈达德的任务运营。位于巴尔的摩的太空望远镜科学研究所由天文学研究大学协会运营，为美国宇航局进行哈勃科学操作。