Thin‐film event‐based vision sensors for enhanced multispectral perception beyond human vision
Abstract Dynamic detection is crucial for intelligent vision systems, enabling applications like autonomous vehicles and advanced surveillance. Event‐based sensors, which convert illumination variations into sparse event spikes, are highly effective for dynamic detection with low data redundancy. Ho...
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-07-01
|
| Series: | InfoMat |
| Subjects: | |
| Online Access: | https://doi.org/10.1002/inf2.70007 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Dynamic detection is crucial for intelligent vision systems, enabling applications like autonomous vehicles and advanced surveillance. Event‐based sensors, which convert illumination variations into sparse event spikes, are highly effective for dynamic detection with low data redundancy. However, current event‐based vision sensors with simplified photosensitive capacitor structures face limitations, particularly in their spectral response, which hinders effective information acquisition in multispectral scenes. Here, we introduce a two‐terminal thin‐film event‐based vision sensor that innovatively integrates an inorganic oxide p–n junction with the pyro‐phototronic effect, synergistically combining the photovoltaic and pyroelectric mechanisms. This innovation enables spiking signals with a tenfold increase in responsivity, a dynamic range of 110 dB, and an extended spectral response from ultraviolet (UV) to near‐infrared (NIR). With a thin‐film sensor array, these spiking signals accurately extract fingerprint edge features even under low‐light conditions, benefiting from high sensitivity to minor luminance variations. Additionally, the sensors' broadband spiking response captures richer information, achieving 99.25% accuracy in multispectral dynamic gesture recognition while reducing data processing by over 65%. This approach effectively eliminates redundant data while minimizing information loss, offering a promising alternative to current dynamic perception technologies. |
|---|---|
| ISSN: | 2567-3165 |