Chinese scientists develop AI model to push deep-space exploration

BEIJING -- Chinese researchers have developed an artificial intelligence (AI) model for astronomical imaging that significantly enhances scientists' ability to peer into the deepest reaches of the cosmos.

A cross-disciplinary research team from Tsinghua University developed the model, named ASTERIS (Astronomical Spatiotemporal Enhancement and Reconstruction for Image Synthesis), using computational optics and AI algorithms.

According to the findings published on Friday in the journal Science, the model can help extract extremely faint astronomical signals, identify galaxies more than 13 billion light years away and generate the deepest deep-space images ever produced.

Exploring distant, faint celestial objects is crucial to understanding the origin and evolution of the universe. Yet astronomers face a major challenge. Weak signals from remote celestial objects are often obscured by background sky noise and thermal radiation from telescopes.

The study shows that applying the model's "self-supervised spatiotemporal denoising" technique to data from the James Webb Space Telescope (JWST) extends observational coverage from visible light at around 500 nanometers to the mid-infrared at 5 micrometers. It also increases the detection depth by 1.0 magnitude, effectively enabling the telescope to detect objects 2.5 times fainter than previously possible.

Using the model, the team identified more than 160 candidate high-redshift galaxies from the "Cosmic Dawn" period, roughly 200 million to 500 million years after the Big Bang, tripling the number of discoveries using previous methods, according to Cai Zheng, associate professor at Tsinghua's Department of Astronomy and a member of the research team.

Researchers said the AI model can decode massive volumes of space telescope data and is compatible with multiple observational platforms, giving it the potential to become a universal deep-space data enhancement platform.

Traditional noise-reduction techniques rely on stacking multiple exposures and assume noise is uniform or correlated. In reality, deep-space noise varies across both time and space. ASTERIS addresses this by reconstructing deep-space images as a 3D spatiotemporal volume.

Through "photometric adaptive screening mechanism," the model identifies subtle noise fluctuations and distinguishes them from the ultra-faint signals of distant stars and galaxies.

"Overall, I think this is a very relevant piece of work that can have an important impact across astronomy," one reviewer of the research said.

Faint celestial objects obscured by light noise in astronomical observations can be reconstructed with high fidelity, said Dai Qionghai, professor at Tsinghua's Department of Automation.

Looking ahead, researchers expect the technology to be deployed on next-generation telescopes to help address major scientific questions concerning decoding dark energy, dark matter, cosmic origins and exoplanets.