A new camera could allow autonomous vehicles to detect hazards, other cars and people three times farther away than the color cameras currently being used.
A team of researchers, inspired by the vision system of the mantis shrimp, has developed a low-cost camera that could help improve the ability of autonomous vehicles to identify possible hazards in challenging imaging conditions.
The new camera, which features a dynamic range—a measurement of the brightest and darkest areas a camera can capture simultaneously—about 10,000 times higher than current commercial cameras, can detect the polarization of light. The unique properties enable the camera to see better in difficult driving conditions, like the transition from a dark tunnel into bright sunlight or during hazy or foggy conditions.
“In a recent crash involving a self-driving car, the car failed to detect a semi-truck because its color and light intensity blended with that of the sky in the background,” research team leader Viktor Gruev of the University of Illinois at Urbana-Champaign, said in a statement. “Our camera can solve this problem because its high dynamic range makes it easier to detect objects that are similar to the background and the polarization of a truck is different than that of the sky.”
Mantis shrimp have a logarithmic response to light intensity that make the sea creatures sensitive to a high range of light intensities. This allows the shrimp to perceive both very dark and very bright elements within a single scene.
The researchers tweaked the way the camera’s photodiodes convert light into an electrical current by operating the photodiodes in a forward bias mode rather than a traditional reverse bias mode to change the electrical current output from being linearly proportional to the light input to having a logarithmic response similar to the shrimp.
The researchers also mimicked how the shrimp integrates polarized light detection into its photoreceptors by depositing nanomaterials directly onto the surface of the imaging chip that contained the forward biased photodiodes.
“These nanomaterials essentially act as polarization filters at the pixel level to detect polarization in the same way that the mantis shrimp sees polarization,” Gruev said.
Additional processing steps were developed to clean up the images and improve the signal to noise ratio.
The team tested the new cameras with different light intensities, colors and polarization conditions in the lab and field.
“We used the camera under different driving lighting conditions such as tunnels or foggy conditions,” Tyler Davis, a member of the research team, said in a statement. “The camera handled these challenging imaging conditions without any problems.”
The researchers are currently working with an air bag manufacturing company to examine whether the camera can be used to better detect objects to either avert collisions or signal to deploy the air bag a few milliseconds earlier.
Along with self-driving cars, the researchers are exploring using the cameras to detect cancerous cells, which exhibit a different light polarization than normal tissue, and to improve ocean exploration.
“We are beginning to reach the limit of what traditional imaging sensors can accomplish,” Missael Garcia, first author of the paper, said in a statement. “Our new bioinspired camera shows that nature has a lot of interesting solutions that we can take advantage of for designing next-generation sensors.”