Intelligent Optoelectronic Devices for Next-Generation Artificial Machine Vision

Machine vision technology is a thriving interdisciplinary subject to simulate human vision-related intelligent behaviors, which has played an increasingly important role in various fields, especially in artificial intelligence and big-data eras. However, the existing silicon-based machine vision technology (s-MVT) has been plagued by complex and inefficient von Neumann architecture, thus gradually becoming overwhelmed when handling high-throughput parallel tasks. The emerging bionic machine vision technology (b-MVT) with new materials, working principles, and architectures are attracting ever-increasing attention, which simulates the structure-function relationship of the human visual system and shows great potential in processing optical information efficiently with low power consumption. In this review, the recent developments achieved in the b-MVT are summarized, with a particular emphasis on material selection, device fabrication and performance, as well as several popular neural networks for back-end image recognition. A summary and critical challenges of the b-MVT are also discussed.