ScholarWorks@중앙대학교

초록

As demand for intelligent machine vision systems grows across diverse fields, neuromorphic computational approaches such as image preprocessing have become crucial for efficient visual processing. Inspired by the biological retina, in-sensor computational systems offer intrinsic efficiency but have largely depended on multi-wavelength strategies for optoelectronic contrast enhancement. Here, retina-mimetic Al2O3-encapsulated tellurium nanowire in-sensor neuromorphic transistors are introduced for image preprocessing. Interestingly, these devices exhibit gate-tunable positive and negative photoconductivity under monochromatic illumination, closely resembling retinal visual processing, such as the bidirectional excitatory/inhibitory behavior. The trap-assisted recombination model, governed by gate polarity, elucidates this gate-bias tunability. Furthermore, bio-inspired image preprocessing under varying noise levels resulted in substantial improvements in recognition, approximate to 93.01% for the weak-noise and 69.01% strong-noise case, highlighting the promise of chalcogenide nanowire devices for highly biomimetic artificial vision systems.

키워드