A near infrared sensing reservoir computing system for entity target recognition

State-of-the-art reservoir computing (RC) systems rely on preprocessed images as the input layer’s signal source, lacking the capacity to directly identify physical targets. In this study, we develop a near-infrared sensor RC system tailored for direct physical target detection. This system integrates a lead sulfide quantum dots (PbS QDs)-based near-infrared sensor array, zinc oxide (ZnO)-based memristors, and peripheral circuits. The PbS QDs sensor array serves as the signal source of the input layer of the RC system, capturing image information through near-infrared light and converting it into a temporal signal that is input into a reservoir composed of ZnO memristors. Memristors exhibit robust resistance switching stability and synaptic characteristics, enabling superior reservoir state separation within RC systems. The results demonstrate that the recognition accuracy for 5 × 4 digital images exceeds 88 %. This system provides a promising way for the future development of intelligent infrared cameras and shows its potential in complex target recognition.