Jiefang——Bionic Vision Sensor

The ability to detect objects in low-light environments (e.g., at night, in dark places, or in foggy conditions) can greatly improve the reliability of autonomous vehicles and mobile robotics systems. However, the most widely used computer vision methods have been found to perform poorly in low light.

Photodiode

Researchers at the Hong Kong Polytechnic University have proposed a new type of biomimetic vision sensor that can adapt to the spectral characteristics of the environment it captures and is based on an array of photodiodes placed next to each other.

The main goal of this research is to design a visual sensor that can outperform other sensors in identifying objects in environments with strong light interference and smoke or fog in the air.

The visual sensor they designed is based on photodiodes arranged in series, consisting of variable junctions with different spectral sensitivities. Specifically, the shallow junction consists of TiO2/Sb2Se3, while the deep junction consists of Sb2Se3/Si. The choice of these two junctions can be controlled by an external bias voltage.

Choosing a shallow junction in the sensor increases the sensitivity of the sensor to short wavelength light, while choosing a deep junction increases the sensitivity of the sensor to long wavelength light. Therefore, the unique design allows its photodiode to be tuned to match the broadband visible light spectrum or the narrowband near-infrared light spectrum.

The spectral adjustment process takes only tens of microseconds, which is comparable to the frame rate of state-of-the-art high-speed cameras.

When applied to the vision systems of autonomous vehicles, the approach was found to provide glare protection with low latency and low power consumption, without the need for additional optical add-ons or complex computational algorithms.

resource

The practical implications of such in-sensor spectral adaptation approaches are significant because they eliminate the need for optical add-ons or complex algorithmic resources.

A new sensor developed by a team of researchers could improve the capabilities of robotic systems, specifically improving their ability to recognize objects in different lighting conditions, without increasing energy consumption.