Bio-inspired Imaging Systems Based on Curved Image Sensors and Soft Neuromorphic Optoelectronic Devices
Restricted (Penn State Only)
- Author:
- Lu, Yuntao
- Graduate Program:
- Engineering Science and Mechanics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 06, 2024
- Committee Members:
- Cunjiang Yu, Chair & Dissertation Advisor
Michael Lanagan, Major Field Member
Mark Horn, Outside Field Member
Mehdi Kiani, Outside Unit Member
Albert Segall, Program Head/Chair - Keywords:
- Bio-inspired imaging
Curved image sensors
Neuromorphic image sensors
Optoelectronic synapses - Abstract:
- Biological visual systems can efficiently perform complicated tasks including image sensing, signal processing, and image recognition. Inspired by biological visual systems, recent years have witnessed an increased interest and significant development in various advanced bio-inspired artificial imaging systems. Artificial imaging systems take inspiration from biology in several different ways. For image acquisition, a curved image sensor that mimics the curvilinear shape of biological retina can provide many advantages such as reduced optical aberration, simplified optical system design, and a wide field of view compared with traditional planar image sensors. For image data processing, neuromorphic devices such as artificial synapses have been developed to achieve more efficient signal processing and reduced redundant modules compared with traditional von-Neumann computing systems. The combination of curved image sensor and neuromorphic imaging can integrate low-aberration image sensing and signal processing into a single device, further enhancing the system-level efficiency. In this dissertation, we first introduce a curvy and shape-adaptive imager with a high pixel fill factor by transferring an ultrathin, stretchable, kirigami-structured silicon photodetector array onto curvy surfaces through a conformal additive stamp printing technique. The curvy, shape-adaptive imager can match its curvature with the focal plane of a single plano-convex lens at different object distances by simultaneously tuning the focal distance of the lens and the curvature of the imager, thus allowing objects at various distances to be imaged while maintaining low optical aberration. Next, we present an artificial optoelectronic synapse that integrates silicon photovoltaic cells and an electric-double-layer synaptic transistor. The optoelectronic synapse exhibits high synaptic facilitation through a unique cascaded synaptic signal transmission mechanism. Moreover, based on this device architecture, a curvy neuromorphic image sensor array that simultaneously mimics the curved geometry and neural signal transmission of the human retina is constructed to demonstrate the visual information sensing and contrast enhancement capability owing to the enhanced synaptic facilitation behavior. In the last section, we focus on a global electrolyte-gated metal oxide transistor array that mimics the highly interconnected biological neural network for spatial signal processing. This device can perform multiply-accumulate operations in an analog manner through global electrolyte gating and ion accumulation with a simple device structure and minimal hardwire connections. We further use this device design to build a flexible neuromorphic image sensor, which can simultaneously perform image sensing and near-sensor image processing.