Researchers have developed a camera that can capture 3D information about objects in a scene with a single exposure. Industrial part inspection, gesture recognition, and collecting data for 3D display systems are some of the things the camera could be used for.
A research team leader from the University of California, Davis said that they consider their camera lensless because it replaces the bulk lens used in conventional cameras with a thin, lightweight microlens. "Because each microlens can observe objects from different viewing angles, it is possible to acquire 3D information from objects partially obscured by objects closer to the camera."
The new 3D camera was described in the journal by the authors. 3D images can be produced in real time because the camera learns from previous data.
The 3D camera could be used to give robots 3D vision, which could help them navigate 3D space or enable complex tasks. It could also be used to provide content for 3D displays used in gaming, entertainment or many other applications.
The camera learns.
The researchers created a compact microscope that could image 3D structures for medical applications. "We built a microscope using a microlens array and thought that a similar concept could be applied to images of macroscopic objects."
The individual lens in the new camera allows it to see objects from a variety of angles. Although other research groups have developed cameras that are based on a single-layer microlens array, it has been difficult to make them practical because of the lengthy process of calibrating and reconstructing them.
To make a more practical 3D camera, the researchers looked at the microlens array and the reconstruction algorithm in the same way. The microlens array is made from just 12 millimeters in diameter and contains 37 small lenses. A highly efficient artificial neural network learns how to map information from the image back to the objects in a scene, which is the basis of the reconstruction method they developed.
The underlying mechanism of many existing neural networks is difficult to explain. The neural network is based on a model. The learning process is made easier by this.
At a very high speed, it can reconstruct images containing objects that are at different distances away from the camera. The new camera can be used to map the 3D locations and outlines of objects.
Seeing things.
The researchers did numerical simulations to verify the camera's performance. The camera's ability to perform 3D images at different depths was tested. The reconstruction could be focused on different depths. A depth map was created by the camera.
The final demonstration showed that the camera could image objects behind the obstacles. This is the first demonstration of using a camera that does not have a lens.
The researchers are trying to reduce artifacts, or errors, that appear in the 3D reconstructions. The miniaturization of the device footprint would make it more portable and enable more applications.
"Our lensless 3D camera uses a new approach that combines hardware and software to achieve desired images and quality." The recent development of low-cost, advanced micro-optics manufacturing techniques as well as improvements in machine learning and computational resources will enable many new images with advanced functions.
More information: Feng Tian et al, Learned lensless 3D camera, Optics Express (2022). DOI: 10.1364/OE.465933 Journal information: Optics Express
All Rights Reserved © 2024
