Lensless camera could help robots see in 3D

This is the claim of a team at the University of California, Davis, who see the lensless camera finding applications in industrial part inspection, gesture recognition and collecting data for 3D display systems.

“We consider our camera lensless because it replaces the bulk lenses used in conventional cameras with a thin, lightweight microlens array made of flexible polymer,” said research team leader Weijian Yang from UC Davis. “Because each microlens can observe objects from different viewing angles, it can accomplish complex imaging tasks such as acquiring 3D information from objects partially obscured by objects closer to the camera.”

In Optics Express, Yang and first author Feng Tian, a doctoral student in Yang’s lab, describe the new 3D camera. Because the camera learns from existing data how to digitally reconstruct a 3D scene, it can produce 3D images in real time.

“This 3D camera could be used to give robots 3D vision, which could help them navigate 3D space or enable complex tasks such as manipulation of fine objects,” Yang said in a statement. “It could also be used to acquire rich 3D information that could provide content for 3D displays used in gaming, entertainment or many other applications.”

The individual lenses in the new camera allow it to see objects from different perspectives, which provides depth information. Other research groups have developed cameras based on single layer microlens arrays, but it is said to have been difficult to make them practical because of extensive calibration processes and slow reconstruction speeds.

To make a more practical 3D camera for macroscopic objects, the researchers considered the microlens array and the reconstruction algorithm together rather than approaching them separately. They custom designed and fabricated the microlens array, which contains 37 small lenses distributed in a circular layer of polymer that is 12mm in diameter. The reconstruction algorithm they developed is based on an artificial neural network that learns how to map information from the image back to the objects in a scene.

“Many existing neural networks can perform designated tasks, but the underlying mechanism is difficult to explain and understand,” said Yang. “Our neural network is based on a physical model of image reconstruction. This makes the learning process much easier and results in high quality reconstructions.”

Once the learning process is complete, it can reconstruct images containing objects that are at different distances away from the camera at a very high speed. The new camera doesn’t need calibration and can be used to map the 3D locations and spatial profiles of objects.

After performing numerical simulations to verify the camera’s performance, the researchers performed 2D imaging that showed perceptually satisfying results. They then tested the camera’s ability to perform 3D imaging of objects at different depths. The resulting 3D reconstruction could be refocused to different depths or distances. The camera also created a depth map that agreed with the actual object arrangement.

“In a final demonstration we showed that our camera could image objects behind the opaque obstacles,” said Yang. “To the best of our knowledge, this is the first demonstration of imaging objects behind opaque obstacles using a lensless camera.”

The researchers are currently working to reduce artifacts in the 3D reconstructions and to improve the algorithms to gain higher quality and speed. They also want to miniaturise the overall device footprint so it could fit into a cellphone.