Hemispherical Bionic Eye Integrates Full-Color Imaging and Wide-Angle Motion Detection

Researchers at The Hong Kong University of Science and Technology have engineered a high-resolution bionic eye featuring an aberration-corrected field of view exceeding 160 degrees, full-color capability, and on-sensor event-driven motion detection. Described in a study published in Nature Materials, the bio-inspired vision system mimics the naturally curved geometry of biological retinas to overcome standard optical trade-offs inherent to rigid, flat image sensors. By consolidating light detection, optical aberration handling, and motion recognition into a unified hardware platform, the development demonstrates a practical architecture for compact, low-bandwidth machine vision in robotics and autonomous navigation.

Conventional machine vision systems typically rely on planar image sensor arrays that require bulky multi-element optical lenses and demanding external computing routines to correct optical aberrations. These optical configurations inherently add significant mechanical bulk, constrain the effective field of view, and draw substantial operational power. While curved focal planes offer an elegant biomimetic solution capable of supporting expansive wide-angle vision within miniature footprints, their fabrication has historically been hindered by the physical limits of traditional lithography, which is universally tailored for planar semiconductor wafers rather than highly curved surfaces.

To overcome this long-standing manufacturing obstacle, the HKUST research team pioneered an advanced 3D laser-assisted fabrication methodology capable of structuring dense arrays of functional, light-responsive pixels directly across a hemispherical substrate. This fabrication breakthrough allowed the team to construct a hemispherical tandem artificial retina packing 367,500 sensing pixels at a density of 1,905 pixels per inch. The resulting resolution marks a substantial technological scaling jump from the team’s initial proof-of-concept bionic eye demonstrated in 2020, which housed only 100 functional pixels on a curved surface.

The newly fabricated tandem artificial retina architecture incorporates distinct blue-, green-, and red-sensitive layers arranged sequentially across the curved substrate. This multilayered design provides continuous full-color visual recording across a broad operational spectral band ranging from 300 to 800 nanometers. Simultaneously, the hemispherical focal geometry ensures aberration-corrected imaging across a wide visual span exceeding 160 degrees, eliminating the heavy multi-lens stacks and post-processing algorithms otherwise required to rectify peripheral distortion in wide-angle planar optics.

Beyond standard optical capture, the tandem retina design natively performs in-sensor, event-driven motion detection. Standard digital video systems rely on frame-based image capture, continuously acquiring, buffering, and transmitting whole frames to downstream processors even when the scene contains little or no relative movement. In contrast, the HKUST bionic eye senses and processes dynamic motion-related information directly at the pixel level, bypassing the requirement to route continuous sequences of complete visual frames to an external computing unit.

Laboratory evaluation demonstrated that this in-sensor, event-driven processing framework slashed data bandwidth requirements by more than 99.95 percent compared to traditional frame-based visual recording methods under tested operational conditions. Despite this massive reduction in the volume of transmitted sensory data, the system maintained an evaluated motion recognition accuracy of 98.6 percent. This sharp decrease in visual data overhead addresses one of the primary bandwidth and computational bottlenecks constraining autonomous robotic platforms.

The research, directed by corresponding author Prof. FAN Zhiyong alongside lead author Dr. LONG Zhenghao and collaborator Prof. ZI Yunlong, illustrates the viability of executing real-time environmental tracking within constrained energy and volumetric budgets. Looking ahead, the researchers anticipate that further refinements to the 3D laser fabrication process could enable next-generation curved vision sensors carrying over one million light-sensitive pixels, advancing practical deployments in compact autonomous robots and machine vision.

The manufacturing innovation addresses a fundamental limitation in device microfabrication on non-planar surfaces. Conventional lithographic processes were designed primarily for planar substrates, creating severe technical obstacles to patterning high-density pixel arrays across steep spherical contours. By applying their three-dimensional laser fabrication strategy directly to hemispherical substrates, the team established a workable pathway for building dense arrays of light-sensitive pixels on curved geometries. The team’s earlier bionic eye prototype, documented in a 2020 paper, demonstrated the initial operating principle but was restricted to only 100 pixels. The current system represents an increase of several orders of magnitude to 367,500 pixels, directly demonstrating scalable lithography-free patterning on three-dimensional hemispherical substrates.

The study, published in Nature Materials under the title “An aberration-corrected bionic eye for full-colour vision and event-driven motion detection,” was carried out through an interdisciplinary institutional collaboration. The experimental investigation was led by Prof. Fan Zhiyong, who holds appointments as Chair Professor in both the Department of Electronic and Computer Engineering and the Department of Chemical and Biological Engineering, and who serves as Director of the State Key Laboratory of Displays and Opto-Electronics at HKUST. Prof. Fan acted as the study’s sole corresponding author, directing the project alongside first author Dr. Long Zhenghao, a former doctoral student and postdoctoral researcher at HKUST who is now an Assistant Professor at Fudan University. The research was conducted in collaboration with Prof. Zi Yunlong from HKUST (Guangzhou).

Co-authors on the research team included Research Assistant Professor Dr. Swapnadeep Poddar and doctoral students Wang Kaichen, Sun Xiaofei, Ye Zilong, Li Jianye, and Shivam Kumar. While the current prototype achieved motion detection within the sensor itself to eliminate full-frame transmissions, further development will focus on scaling fabrication techniques toward next-generation devices with over one million light-sensing pixels for potential deployment in autonomous driving systems and robotics.

Subject of Research: Hemispherical tandem artificial retina for full-color imaging and event-driven motion detection

Article Title: HKUST develops high-resolution bionic eye

Article References: HKUST develops high-resolution bionic eye. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: bionic eye, artificial retina, curved image sensors, 3D laser fabrication, event-driven motion detection, full-color imaging, machine vision, robotics, autonomous systems, optical aberration correction, wide-angle vision, in-sensor computing, Nature Materials, HKUST, neuromorphic engineering, optical sensors