A Biomimetic Fisheye for High-Accuracy Underwater Recognition Enabled by Oxide Semiconductor Retina

Area of Science:

• Materials Science and Engineering
• Biomimetic Systems
• Optoelectronics

Background:

• Underwater imaging faces challenges due to scattering and complex backgrounds.
• Machine vision requires high sensitivity, broadband response, wide field of view, and efficient preprocessing for marine exploration.

Purpose of the Study:

• To demonstrate an oxide-semiconductor-based biomimetic fisheye vision system for high-accuracy underwater recognition.
• To develop an artificial retina with enhanced sensitivity and computational capabilities for marine environments.

Main Methods:

• Designed a sandwich-structured phototransistor array (n-ITZO/p-SnO/n-ITZO) with high-mobility ITZO and narrow-bandgap SnO.
• Optimized the 3D curvature of the retina to correct optical aberrations for wide-field-of-view imaging.
• Integrated the artificial retina with algorithms for a hardware-based underwater recognition framework.

Main Results:

• The phototransistor exhibited a broadband response (UV to NIR) with high detectivity (>10^11 cm Hz^1/2 W^-1).
• The system demonstrated excellent neuromorphic capabilities, including short-term memory and temporal sensitivity.
• Achieved an average underwater fish recognition accuracy of 90.86%, robust to Gaussian and salt-and-pepper noise.

Conclusions:

• The developed biomimetic fisheye vision system offers a compact and efficient hardware solution for underwater visual tasks.
• This technology advances marine resource exploration through high-accuracy, noise-resilient underwater recognition.
• The artificial retina serves as a physical reservoir, supporting in-sensor computing for next-generation underwater visual systems.