Bioinspired artificial vision system based on photoelectric memristors

Traditional artificial vision systems are constrained by the von Neumann architecture, which segregates the components responsible for image perception, memory, and processing. This segregation results in low integration, high energy consumption, and inefficient processing. In contrast to traditional vision systems, bio-inspired neuromorphic vision systems utilize photoelectric memristors to integrate sensing, memory, and computation, enabling parallel processing of visual information. These systems can not only emulate the retina direct response to light signals to perform image preprocessing tasks, such as noise reduction and contrast enhancement, but also replicate the image recognition and classification functions of the brain visual cortex. Additionally, they offer a reliable hardware platform to enable progress in complex artificial vision technologies. This review primarily summarizes the research progress in artificial visual systems based on photoelectric memristors, offering an overview of their device structure, working mechanisms, material and device design, as well as their applications and challenges in artificial vision. This work is intended to analyze new technologies and their future outlook, focusing on how they may fuel innovation and facilitate extensive adoption.