HKUST Laser-Etches 367,500-Pixel Bionic Eye Sensor Cutting Data 99.95%

HKUST researchers engineered a hemispherical bionic eye that reduces transmitted data volume by 99.95% by integrating movement detection into the pixel array.

Researchers at the Hong Kong University of Science and Technology (HKUST) have engineered a hemispherical bionic eye capable of high-density vision processing.

By utilizing laser-etched pixels on a curved sensor, the device eliminates the need for bulky corrective lenses, providing a 160-degree field of view with 367,500 individual pixels.

This development, detailed in Nature Materials on November 13, 2026, changes machine vision architecture, according to Science et vie. By integrating movement detection directly into the pixel array, the system reduces transmitted data volume by 99.95% while maintaining a 98.6% motion recognition accuracy, offering a potential path to lower power consumption for autonomous systems.

  • Data Efficiency: The 99.95% reduction in data transmission mitigates bandwidth bottlenecks in real-time robotics and autonomous vehicle (AV) processing units.
  • Form Factor Optimization: Eliminating lens stacks reduces the physical footprint of sensor modules, potentially lowering manufacturing costs and complexity for hardware integrators.
  • Scaling Roadmap: Research lead Fan Zhiyong is currently targeting a transition toward a million-pixel density, a necessary milestone for commercial-grade industrial adoption.

Traditional image sensors rely on flat-plane architectures, which require complex lens stacks to correct for spherical aberration. These components add weight and energy consumption to robotic systems.

The HKUST team’s approach mimics biological morphology, allowing the device to capture a 160-degree field of view without the distortion typically associated with wide-angle flat lenses. According to the research published in Nature Materials, the sensor achieves a resolution of 1,905 points per inch.

The technical architecture replaces conventional color filter mosaics with three stacked, light-sensitive layers tuned to red, green, and blue. This stack processes light across a spectrum ranging from near-ultraviolet to near-infrared without the traditional reliance on external corrective optics. The result is a system that processes motion at the pixel level rather than through full-frame comparisons.

Metric Performance Specification
Pixel Count 367,500
Data Reduction 99.95%
Motion Detection Accuracy 98.6%
Field of View >160 Degrees

By performing detection at the sensor level, the HKUST architecture addresses the high energy costs that have historically plagued high-performance machine vision.

However, transitioning from a laboratory prototype to a mass-manufactured component is difficult. While the 2020 iteration of this technology featured only 100 pixels, the current 367,500-pixel density shows an exponential increase in capability. The primary question remains whether this laser-etching process can be adapted for high-volume production lines at a cost point competitive with current CMOS sensor manufacturing.

The research team has not yet disclosed a timeline for commercial pilot programs or partnerships with hardware manufacturers. Fan Zhiyong continues to oversee development toward the target of one million pixels, a density threshold required to compete with human-equivalent visual acuity in complex environments.

The Hong Kong University of Science and Technology, which serves as the hub for this research, was founded in 1991 as the territory’s first research university. Occupying a 60-hectare seaside site in Tai Po Tsai, the institution was built on land originally earmarked for a British Army garrison. It holds the distinction of being the first university in the region to be established without any precursory existence.

Disclaimer: The information provided in this article is for educational and informational purposes only and does not constitute financial advice.

Photo of author

Newsroom

Newsroom at Archyde.

Indonesian STIP Cadets Win Two Golds at World Robot Games Japan 2026