기초과학연구원(IBS, 원장 노도영) 나노입자 연구단 김대형 부연구단장(서울대 화학생물공학부 교수) 연구팀과 송영민 광주과학기술원(GIST) 교수팀이 공동으로 새 눈의 구조와 기능을 모방한 물체 감지 특화 카메라를 개발하며, 물체 감지가 필요한 무인로봇, 자율주행차, 드론에서 적용이 증가할 것으로 기대된다.

▲Schematic diagram and measurement results of a camera that mimics the new eye developed by the research team
Application to unmanned robots, autonomous vehicles, and drones that require object detection
A new camera has been developed that resembles the eyes of an eagle, capable of capturing moving prey from several kilometers away.
The research team of Kim Dae-hyung, deputy research director of the Nanoparticle Research Group at the Institute for Basic Science (IBS, President Noh Do-young) and professor of chemical and biological engineering at Seoul National University, jointly developed an object detection camera that mimics the structure and function of a bird's eye with the team of Professor Song Young-min at the Gwangju Institute of Science and Technology (GIST).
The eyes of animals are optimized for their habitat and survival environment. The eyes of birds such as eagles have evolved to have a deep and narrow fovea on the retina to suit their survival environment of living in high trees. The deep and narrow fovea is advantageous for magnifying distant objects.
The central part of the bird's eye is densely filled with color-detecting cone cells, which allow them to perceive objects more clearly. While human eyes can only detect visible light (red, green, and blue), bird eyes can also detect ultraviolet light. Thanks to this, birds can obtain visual information that humans cannot see, and can effectively recognize objects even in complex and dynamic environments.
The IBS research team designed a new camera specialized in object detection, inspired by the structure and function of bird eyes. The developed camera consists of an artificial fovea and a multispectral image sensor capable of detecting visible light and ultraviolet light.
First, through collaboration with Professor Song Young-min’s team at GIST, we created an artificial fovea that mimics the central fovea structure of birds. Through optical simulation, we devised an optimal design that can magnify distant objects without image distortion.
Afterwards, a multi-wavelength image sensor was created using perovskite materials with excellent electrical and optical properties. After creating an optical sensor using four types of perovskite materials that absorb different wavelength ranges, they stacked them vertically to implement a sensor that can distinguish colors without a color filter.
“Our research team also developed a new transfer process for producing a multi-wavelength image sensor,” said co-first author Jin-Hong Park, a researcher at the IBS Center for Nanoparticle Research. “Combining this with the perovskite patterning technology developed in previous research, we were able to complete a sensor that can detect not only visible light but also ultraviolet rays without a filter.”
Conventional cameras that use zoom lenses to magnify objects have the disadvantage of not being able to recognize the periphery of the magnified object.
On the other hand, the object detection specialized camera developed by the research team magnifies the object in the central part of the field of view while also providing a peripheral view. As a result, the movement of the object can be detected more sensitively based on the difference between the two fields of view. In addition, since it detects visible light and ultraviolet rays without a filter, the visual information becomes more diverse, and there are also advantages in reducing the process cost and weight.
The research team verified the object recognition and motion detection capabilities of the developed camera through simulation. In terms of object recognition ability, the new camera (confidence score 0.76) showed a confidence score that was about twice as high as that of the existing camera system (confidence score 0.39). The rate of change in movement was also 3.6 times higher than that of the existing camera system, confirming that it can detect movement more sensitively.
Kim Dae-hyung, deputy head of the research team that led the research, said, “Birds’ eyes have evolved to have a structure that is advantageous for quickly and accurately recognizing distant objects even when flying at high altitudes.” He added, “The camera developed by our research team can be applied to unmanned robots and autonomous vehicles that require object detection capabilities, and we expect it to be particularly effective when installed in drones that operate in environments similar to those of birds.”
The research results were published on May 30 (Korean time) in the online edition of 'Science Robotics (IF 25.0),' the world's top academic journal in the field of robotics.