Imaging and Classification of Objects Through Random Diffusers and Occlusions at the Speed of Light
Abstract
Imaging and sensing through diffusive or occluding media present a very challenging problem. Various application areas and fields su,ch as biomedical imaging, astronomy, atmospheric sciences, oceanography, security/defense, robotics, autonomous driving, among other,s, will significantly benefit from a method that can rapidly image and classify objects through random diffusers and opaque occlusio,ns. All the existing body of solutions that have been demonstrated for computational imaging and classification of objects through d,iffusers or occlusions have been based on digital computers and algorithms, and therefore are limited in their speed, energy consump,tion, portability and scalability, especially in resource-constrained settings that demand extremely fast and low-power response. Th,is proposal focuses on computational imaging without a digital computer and aims to design and demonstrate a computer-free, all-opti,cal platform that will image and sense through diffusive media or opaque occlusions, achieving object recognition and image reconstr,uction at the speed of light propagation. Unlike digital image reconstruction or classification methods that use digital computers a,nd algorithms, we propose an all-optical method using a passive platform (i.e., no power consumption, except for the illumination li,ght) composed of a set of transmissive diffractive surfaces that will be jointly optimized to all-optically recognize/classify or re,construct objects behind an unknown diffuser or opaque occlusion. These all-optical diffractive processors and the resulting passive, networks can operate at different parts of the electromagnetic spectrum by accordingly scaling the diffractive features proportiona,l to the wavelength of light and therefore can be used with various light sources and detectors that are of interest to ONR, includi,ng but not limited to those in the visible, infrared, terahertz and millimeter-wave regions of the spectrum. These proposed platform,s will be able to infer the class/type of unknown objects almost instantaneously and without the need for any digital computer or ex,ternal computing power, making them ideal for various naval applications such as extremely high-speed and low-power surveillance and, automated monitoring of potential threats in field-settings or on naval ships. It is also conceivable to create a powerful network, of task-specific imagers and sensors/classifiers that can operate at different parts of the electromagnetic spectrum using various, replicas of these passive and compact all-optical imagers/classifiers that are constantly processing and searching for specific tar,gets, potentially providing new capabilities for naval operations.(Approved for Public Release)
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 10, 2021
- Source ID
- N000142212016
Entities
People
- Aydoğan Özcan
Organizations
- Office of Naval Research
- United States Navy
- University of California, Los Angeles