Rational design of a vision fusion system with visible and near-infrared spectral integration for improved...

With the rapid advancements in autonomous driving, pure vision-based solutions have garnered significant attention. However, existing vision sensors are limited by their specific spectral operating ranges and the complexity of processing hybrid optical/electrical signals. In this study, we present a fully circuit-emulated vision system that employs a vision fusion solution for autonomous driving, integrating image sensing, fusion, edge extraction, and decision-making functionalities. This system utilizes vision sensors featuring an Al₂O₃/two-dimensional Ruddlesden-Popper perovskite (2D PVK) heterostructural dielectric and MoS₂/black phosphorus (BP)/MoS₂ heterostructural channel, which exhibits persistent nonvolatility and fully light-tunable positive and negative photoresponses when exposed to 1064 nm and 532 nm light, respectively. Notably, when combined with edge extraction circuit design, our vision system achieves all-day visual perception with a 99.0% recognition accuracy for driving scenario information. The integration of the fully circuit-emulated vision system with the vision fusion solution enables a more comprehensive and accurate representation of the driving environment.