Hong Kong Polytechnic University successfully develops advanced visual sensor that mimics the adaptability of human eyes

On May 31, the Hong Kong Polytechnic University (PolyU) announced that it had recently achieved a scientific breakthrough with scientists from South Korea’s Yonsei University. The light sensor they developed can emulate or even surpass the human retina’s ability to adapt to different brightness levels, and is expected to equip future driverless vehicles and industrial cameras with human-like vision.

“The new sensor can significantly improve machine vision systems, enabling better performance in a variety of image analysis and recognition tasks,” said Dr Chai Yang, Associate Professor of PolyU’s Department of Applied Physics and Assistant Dean of the School of Applied Science and Textiles, who led the research.

Machine vision systems are typically composed of multiple cameras and computing units that capture and process images to perform tasks such as facial recognition. Such systems require sophisticated circuits and complex algorithms to "see" objects under various lighting conditions. However, such systems rarely have sufficient performance to process large amounts of visual information in real time, which is difficult to compare with the human brain.

The new bioinspired sensor developed by Dr. Chai's team may provide a solution to the above problems: the sensor directly adapts to light and darkness, reducing reliance on background calculations. The new sensor imitates the human eye's ability to adapt to different brightness levels, just as the naked eye can accurately identify various objects under extremely dark to extremely bright lighting conditions.

Image source: Hong Kong Polytechnic University

Dr. Chai explained: "The human pupil helps regulate the amount of light entering the eye, while the retinal cells are mainly responsible for adapting to brightness." The natural light intensity range is as wide as 280 decibels, while the adaptation range of traditional silicon-based sensors is only 70 decibels. The new sensor developed by Dr. Chai's team has an effective range of up to 199 decibels, which is impressive. In contrast, the human retina can adapt to environments with brightness ranging from strong light to weak light, with a range of about 160 decibels.

To achieve the above performance, the research team first used molybdenum disulfide, a semiconductor with unique electrical and optical properties, to make a double-layer ultra-thin film with a thickness of nearly atoms, and used it to develop a phototransistor for detecting light; then they introduced a "charge trap state" in the double-layer film to control the ability to detect light.

The "charge trap states" used in phototransistors are actually impurities or defects in the solid crystal structure that are used to restrict the movement of charge. "These trapped states allow light information to be stored and dynamically adjust the optoelectronic properties of the device at a scale as small as each pixel," said the researchers. By controlling the movement of electrons, the trapped states can precisely adjust the amount of electricity conducted by the phototransistor, thereby controlling the device's photosensitivity, that is, its ability to detect light.

Each bionic vision sensor is made up of an array of these phototransistors. They mimic the rods and cones in the human eye, which are responsible for detecting dim and bright light, respectively. As a result, the sensor can detect different objects in a variety of lighting conditions, and can also switch and adapt to different brightness levels, with a range greater than that of the human eye.

Dr. Chai said: “The new sensor can reduce hardware complexity and significantly improve image contrast under different light and dark conditions, thereby providing efficient image recognition.”

This new bionic vision sensor will drive the application of a new generation of artificial vision systems, which will not only be effective in autonomous vehicles and manufacturing applications, but also find other exciting new applications in edge computing and the Internet of Things.