Researchers developed a nanophotonic processor integrating quasi-BIC metasurfaces with MQWs, demonstrating tunable photoresponse for advanced machine vision.
Key Details
- 1Monolithic integration of quasi-bound state metasurfaces with GaAs/AlGaAs multiple quantum wells (MQWs) was achieved.
- 2The system enables vertical-field light absorption in MQWs through a controlled 'leaky mode' in quasi-BIC.
- 3Photoresponse can be tuned nonlinearly by asymmetry or linearly by incident angle and bias voltage.
- 4Prototype neural network implementation processes image contrast and performs edge detection on-chip.
- 5Fabricated devices demonstrated both broadband spectral response and high-density integration feasibility.
- 6Plans include scaling arrays to 1,000+ pixels and enabling real-time object tracking and optically encrypted communication.
Why It Matters
This breakthrough demonstrates hardware-based, efficient, and scalable machine vision technology directly applicable to intelligent imaging systems for radiology AI. It could significantly improve on-chip image preprocessing, neural network implementation, and large-scale integration of sensing and analysis hardware.
Source
EurekAlert
Related News
•EurekAlert
AI Achieves Precision in Generating Radar Images at Specific Attitudes
Researchers introduce a ControlNet-based AI method for precise, attitude-controllable radar image generation of noncooperative targets.
•EurekAlert
RADAR AI Sets New Standard in Abdominal CT Diagnosis
The RADAR vision-language AI model shows expert-level diagnostic ability for abdominal CT across multiple diseases and clinical settings.
•EurekAlert
UCSD Unveils 4D AI Virtual Cell Models to Accelerate Drug Discovery
UC San Diego researchers have developed AI-powered virtual cell models using 4D imaging to predict cellular responses to drugs, potentially expediting drug development.