Disadvantages of existing technology
1. Poor weak light performance of CMOS: It can only perceive 1.0 lux illumination and is difficult to detect weak light signals at night/deep space.
2. Traditional computing has high energy consumption: The "perception processing separation" architecture (such as GPU) consumes 9 orders of magnitude more energy than biological systems.
3. Poor adaptability of neuromorphic devices: low sensitivity to low light, unable to simulate the dark adaptation mechanism of owls.
Article Highlights
1. Optoelectronic decoupling dual-mode design: Simulate owl dark adaptation, enhance photosensitivity with a three terminal transistor structure (insulating polymer coating+light absorption layer), and simulate parallel photon sensing and electrical plasticity.
2. Ultra high low light detection: Active adaptation index of 331, sensing 0.146 nW · cm ⁻ ² low light (3 orders of magnitude higher than CMOS).
3. Stable synaptic plasticity: cyclic LTP/LTD, supporting deployment weights across three types of neural networks with light intensity ranging from 0.146 to 11.7 nW · cm ⁻ ².
4. Brain like system integration: Unmanned aerial vehicle air to ground recognition system, with a recognition accuracy of over 95% at 0.146-11.7 nW · cm ⁻ ².
Application scenarios
1. Night monitoring: City security and border patrol without auxiliary lighting target recognition.
2. Drone search and rescue/exploration: nighttime search and rescue, deep space (moon/Mars) material identification.
3. Precision guidance: target tracking in low light environments of missiles and satellites.
summary
Passive target detection in photon deficient environments is crucial for expanding the capabilities of machine vision in many applications such as precision guidance, intelligent monitoring, and early warning. Here, inspired by owl vision, the author reports a dual-mode synaptic transistor with optoelectronic decoupling mechanism, which can achieve parallel photon perception and electrical plasticity simulation. As a result, the device exhibits a high active adaptation index of approximately 331 and has the ability to perceive weak light intensities as low as 0.146 nW cm ⁻ ². The author also implemented synaptic weight modulation with cyclic stability, exhibiting long-term enhancement and inhibition behavior, and verified the feasibility of deploying weights across three basic artificial neural layers in the light intensity range of 0.146-11.70 nW cm ⁻ ² through adaptive contrast enhancement. The device inspired by owl vision lays the hardware foundation for brain like visual sensors aimed at energy efficiency and low light image processing.