Chinese researchers have developed a new chip that enables real‑time hyperspectral imaging analysis, cutting out bulky servers and huge data files.

Hyperspectral imaging identifies what things are made of. Every material reflects, absorbs and emits light differently across the electromagnetic spectrum. Hyperspectral imaging captures this unique "spectral fingerprint," making it vital for remote sensing and industrial inspection. But the extra spectral dimension comes with a cost: massive data.

The work, led by a team from the Beijing Institute of Technology (BIT), was recently published in the journal Science.

"A single image can take up hundreds of megabytes or even gigabytes, placing huge pressure on back‑end data processing," said Bian Liheng, a tenured professor at BIT.

The team tackled the problem on two fronts. In 2024, they miniaturized the bulky optical system of traditional hyperspectral cameras down to chip level. Now, they've solved the back‑end real‑time computing bottleneck, by mining data correlations to cut redundant calculations, and redesigning the chip's arithmetic units to process data continuously on a single chip.

The breakthrough could transform satellite remote sensing. "When computing is chip‑based, data no longer needs to be transmitted to the ground for processing – satellites can produce results in real time, right on board," Bian said.

This innovation shifts hyperspectral imaging from offline post‑processing to on‑site real‑time computing, opening new possibilities for portable, embedded systems across a wide range of applications.