Robots, drones, and self-driving cars are becoming increasingly embedded into our daily lives, but their sight is far from perfect. Often, they’re stumped by something as simple as sunlight glinting off metal or shadows hiding objects in the dark. While computer vision algorithms have made impressive strides, they still struggle in some situations that humans find trivial to discern. But a team of Korean scientists, led by Professor Young Min Song from the Gwangju Institute of Science and Technology (GIST), may have found a natural fix in an unlikely place: the eyes of cats.
With a little help from evolution, cats have developed a unique set of tools to stay sharp-eyed, day or night. By day, their eyes’ vertical slit pupils narrow to block excessive light, sharpening their focus and reducing glare. When night falls, those pupils open wide to let in more light, and a reflective layer known as the tapetum lucidum boosts their night vision, giving their eyes that characteristic nighttime glow (and weird laser eyes if seen through a camera’s lens).
Inspired by these adaptations, Song’s team has developed a robotic vision system that mimics these feline eye features, enhancing robots’ ability to spot fine details in complex settings and varied lighting.
A Lens Built Like a Cat’s
At the core of the new system is an advanced lens that, like a cat’s slit pupil, filters out distracting light while honing in on key objects.
“Robotic cameras often struggle to spot objects in busy or camouflaged backgrounds, especially when lighting conditions change,” says Song. “Our design solves this by letting robots blur out unnecessary details and focus on important objects.”
This streamlined focus reduces the need for high-powered computer processing, making the system more energy-efficient — a quality increasingly valuable in the world of robotics, especially for drones whose range is still very limited.
The team’s vision system also borrows the tapetum lucidum’s unique design, using a reflective layer behind the photodetector that, much like a cat’s eyes, bounces incoming light back through the lens. This boosts visibility in low light by allowing light to hit the sensor twice, effectively doubling its sensitivity. In early tests, the system outperformed traditional lenses in dim and bright environments alike, excelling in discerning shapes and details where conventional cameras would falter.
From Rescue Operations to Industrial Robots
These developments may improve everything from self-driving cars to search-and-rescue drones. Designed to detect objects in complex, shifting settings, the system could be a game-changer in scenarios where human eyes alone might miss crucial details. That’s because, with this vision system, robots can effectively “see” through cluttered or dimly lit environments.
<!– Tag ID: zmescience_300x250_InContent_3
–>
As Song puts it, “From search-and-rescue operations to industrial monitoring, these cutting-edge robotic eyes stand ready to complement or even replace human efforts in a variety of critical scenarios.”
The team’s work was published in Science Advances.