Metal Oxide-Based Neuromorphic Artificial Visual Perception Devices and Systems for Information Perception, Memory and Processing

Abstract

Vision is recognized as a critical source of environmental perception for humans, with 80% of information sourced visually. The development of sensors that approach or even surpass human visual capabilities has been a continuous pursuit by researchers and is crucial for advancing fields such as autonomous driving and intelligent security. Recently, neuromorphic devices based on metal oxides have become a focal point of research in biomimetic visual perception due to their sensing, storage, and logical computation advantages. This review introduces the distinct characteristics of human vision, traditional image system, and neuromorphic vision system. Then, the classification, material properties, potential mechanisms, and strategies for enhancing stability of metal oxide-based neuromorphic devices are systematically discussed. Finally, the applications of neuromorphic vision systems in image memory and enhancement, motion trajectory, depth perception, vision impairment, and recovery are presented, with a projection on the future development of neuromorphic visual systems. This review aims to provide significant reference and guidance for the future advancement of neuromorphic devices in the fields of biomimetic visual perception, humanoid robots, and autonomous driving.