Energy Efficient Computational Illumination and Imaging
Abstract
Statement of Work:Purchase equipment for the development of novel 3D imaging systems that require low-power to operate and work in high-intensity ambient lighting (e.g. outdoors in the sun).Objective:Purchase equipment for the development of novel 3D imaging systems that are energy-efficient and can work in high-intensity ambient lighting (e.g. outdoors in the sun).Approach:The research supported by this equipment focuses on developing the family of two-layer light modulationarrangements, with one layer controlling illumination and the other controlling a sensor mask. The code sequences for these two programmable masks have so far been constructed heuristically with no regard to energy efficiency. In contrast, the energy efficient codes that the PI derives here transmit more photons to the camera for a given light source power and produce superior images for a given number of photons transmitted. Masked sensors are less energy efficient than mask-less ones because masks may block incident photons. Despite this apparent limitation, programmable masks come with an important advantage: by changing masks and illuminations repeatedly in a single exposure, we can capture images that are impossible to capture with an unmasked sensor in one shot. This generaltechnique expands the capabilities of an imaging system at a cost of reduced energy efficiency. Such a novel imaging approach has many Naval/DoD relevant applications such as seeing through obscurants, obtaining information about the reflecting material, 3D imaging for scene structure inference and object recognition. The PI~s research is currently funded by grants from ONR, Army Robotics CTA and DARPA.Overall Merit and ONR Mission/Relevance:The proposed equipment is relevant primarily to the ONR Autonomy Focus Area, as well as the InformationDominance focus area. The proposed equipment supports funded projects that are in the area of novel imaging systems for enhanced vision and image analysis for use by robots and humans. The proposed equipment supports research in developing novel imaging systems that have many Naval/DoDapplications in surveillance and robotics. The PI and CMU have excellent research and education programs in imaging science and computer vision.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 30, 2016
- Source ID
- N000141612906
Entities
People
- Srinivasa G. Narasimhan
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy