Passive Four-Dimensional Imaging and Recognition in Low Light Levels with Visible Range Image Sensors

Abstract

We propose to investigate novel machine-learning-based passive imaging and classification systems, based on four dimensional (3D spatial plus temporal) integral imaging and operating with CMOS cameras in extremely low light levels. CMOS camera imagery in low light appears as random noise due to the lack of sufficient photons from the scene and is markedly dominated by various sources of camera noise. Our approach to remedy these problems uses advanced machine-learning algorithms and integral imaging architecture, which is optimum for reconstructing 4D images in low light, according to Bayesian algorithms. The results of machine-learning-based passive 4D imaging will be compared to results from long wave IR (LWIR) cameras, which are currently used for low light conditions. The drawbacks of LWIR imaging are: 1) very low spatial resolution (more than an order of magnitude lower than visiblerange cameras), and 2) the higher cost and size of LWIR cameras compared with visible rangecameras.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 31, 2020
Source ID
N000142012690

Entities

People

  • Bahram Javidi

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Connecticut

Tags

Readers

  • Image Processing and Computer Vision.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks