Researchers from Hebei GEO University and Shijiazhuang Tiedao University in China have developed a flexible optoelectronic memristor based on a composite of graphene oxide (GO) and perovskite quantum dots, designed to sense and process images in dimly lit conditions.
Memristors are seen as a key building block for flexible neuromorphic vision systems. Because they can rapidly tune their resistance in response to optoelectronic inputs, they emulate the way biological synapses adjust signal weights - the behavior needed for artificial retina-like sensors that sense and compute in the same place, rather than shuttling data to a separate processor. The team, led by Jingjuan Wang and Lingzhi Tang of Hebei GEO University's College of Information Engineering, combined GO sheets with perovskite quantum dots to create a switching layer whose behavior changes markedly between dark and illuminated conditions.
Under dim illumination, the memristor's conductance range expands, which lets the device filter noise more effectively and, in practice, amplify faint optical signals. Fed low-contrast images, the device outputs clean silhouettes and enables feature extraction for object recognition in low light.
According to the researchers, this device-level tuning carries straight through to system performance: for noisy, low-light images, recognition accuracy improves by more than 10%, the signal-to-noise ratio doubles, and foreground and background signals separate cleanly - all handled locally on the flexible sensor itself, with no additional circuitry required.
The graphene oxide component is central to the device's mechanical durability as well as its electrical behavior. The composite proved highly flexible in testing, continuing to operate reliably after thousands of bending cycles, and its resistance to cracking under deformation makes it a candidate for wearable use - including attachment to curved skin to pick up weak ambient light signals.
Looking ahead, the team plans to pair its flexible memristor arrays with low-power circuits to enable real-time image classification, and to engineer polarization- and wavelength-specific responses in an effort to approach human color vision.
