Description:
- Autonomous vehicles
- Robotics
- Surveillance and security
Abstract
USC researchers bring advanced retinal computations to image sensors with Integrated Retinal Functionality in Image Sensors (IRIS). By leveraging recent advancements in inner retinal circuits, IRIS goes beyond basic luminance adaptation and change detection. It mimics the feature-selective spike trains of retinal ganglion cells (RGCs), enabling two crucial motion computations: Object Motion Sensitivity (OMS) and Looming Detection (LD). With IRIS, cameras can detect moving objects even while in motion, flagging approaching threats, making it ideal for applications such as autonomous vehicles.
Benefits
- Enhanced image sensor capabilities
- Accurate motion and looming detection
- Low-energy, ultra fast operation
- Compatible with existing sensor fabrication processes
Market Application
Neuromorphic image sensors, inspired by the biological retina, have the potential to revolutionize machine vision. Unlike conventional image sensor technology, these sensors can replicate feature-selective computations. By emulating the diverse and specialized circuits of the retina, neuromorphic image sensors offer improved efficiency and performance in machine vision. Key areas of focus are luminance adaptation, enabling high dynamic range imaging, and change detection, identifying object motion, approaching threat detection allowing for high-speed and efficient operation through asynchronous transmission.
Publications
Yin, Zihan, et al. "IRIS: Integrated retinal functionality in image sensors." Frontiers in Neuroscience 17 (2023): 1241691. https://doi.org/10.3389/fnins.2023.1241691
Stage of Development
- Simulation tested
- Available for licensing