MRS Meetings and Events

 

EL05.01.02 2023 MRS Fall Meeting

Bioinspired Compound Eye Camera for High-Resolution and Depth-Based Sensing Applications

When and Where

Nov 27, 2023
11:00am - 11:15am

Hynes, Level 2, Room 203

Presenter

Co-Author(s)

Sehui Chang1,Min Seok Kim1,Gil Ju Lee2,Young Min Song1

GIST1,Pusan National University2

Abstract

Sehui Chang1,Min Seok Kim1,Gil Ju Lee2,Young Min Song1

GIST1,Pusan National University2
Machine vision has advanced to provide high-performance imaging for various fields in electronics and robotics such as smartphones, surveillance, and autonomous vehicles. The capabilities of high resolution, a small form factor, and functional imaging (e.g., depth extraction, object detection, etc.) are of major interest to modern vision platforms. However, conventional camera designs show an intrinsic trade-off between miniaturization and high-performance, including a wide field of view and low aberrations. In particular, a camera design requires multiple lenses to be compatible with a commercial planar CMOS sensor in order to compensate for optical aberrations (e.g., field curvature). This complexity of lens design originates from offsetting against an inherently curved focal plane, which eventually hinders the miniaturization of entire system.<br/>Meanwhile, natural eyes carry out high-performance and functional imaging specialized to their natural habitat, despite having a compact, simple configuration of the ocular structure. For instance, the fish eye provides a wide field of view via a single spherical lens and a curved retina, and the eye of a fiddler crab enables clear imaging both on land and in water via a flat, graded refractive index microlens array. Thus, various animal eyes can inspire the development of enhanced vision systems with novel camera design strategies. Insect compound eye, inter alia, has inspired in various artificial vision systems due to their unique features of wide field of view with low aberrations, high sensitivity to motion, and infinite depth of field. However, most artificial compound eye cameras suffer from two major issues: low resolution caused by the use of a curved photodetector array, or limited field of view caused by the use of a planar image sensor.<br/>Here, inspired by insect compound eyes, we propose a high-resolution compound eye camera for depth-based sensing applications. The artificial compound eye camera consists of a curved microlens array (cMLA), a tapered fiber bundle (TFB), and a commercial CMOS imager. The cMLA was fabricated using flexible polymer, polydimethylsiloxane (PDMS), to achieve the curvature of the MLA through a repeated molding process with a precisely machined flat MLA metal mold and a 3-D printed base mold, enabling wide field of view imaging. To relay sub-images of each microlens in the cMLA onto the planar image sensor plane, we utilized the TFB as a relay optics, which is tapered and polished to form each individual fiber facet perpendicular to its optical axis on its curved and flat sides. The fabricated systems demonstrated high-resolution imaging via the TFB with over 2 million fibers and the commercial sensor, while the entire microlenses of the cMLA provides focused images. Since each individual microlens is matched with ~5 thousands pixels, the acquired image data from the cMLA can provide depth information of objects by calculating disparities among adjacent microimages, while achieving wide field of view imaging. Unlike conventional light field cameras, the proposed artificial compound eye has a curved shape MLA. Therefore, geometrical calibration was conducted to convert the world coordinate into the camera’s spherical coordinate for restoring the visual data in 3-dimensional space. By collecting and processing the image data of angular and distant objects, it is expected that the proposed camera system will pave the way for the development of bioinspired artificial vision specialized in depth sensing applications.

Symposium Organizers

Michael Ford, Lawrence Livermore National Laboratory
Cindy Harnett, University of Louisville
Juejun Hu, Massachusetts Institute of Technology
Seungwoo Lee, Korea University

Publishing Alliance

MRS publishes with Springer Nature