Hyperspectral imaging - status report 2015

Authors:

  • David Bergström
  • Mikael Lundberg
  • David Gustafsson
  • Thomas Svensson
  • Niclas Wadströmer

Publish date: 2016-01-08

Report number: FOI-R--4150--SE

Pages: 35

Written in: Swedish

Keywords:

  • Hyperspectral imaging
  • HSI
  • electro-optical sensors
  • reconnaissance
  • spectral calibration
  • radiometric calibration
  • signature matching
  • anomaly detection
  • material classification
  • Opticks
  • SOCET GXP®
  • HSI-toolbox

Abstract

Aerial and satellite based hyperspectral imaging systems for military applications have been available for a long time. Fundamental applications are target detection and identification of military vehicles, camouflage and man-made materials and terrain classification and mapping. New sensor technologies which are cheap and more compact in size show great promise in providing hyperspectral imaging capabilities to smaller platforms such as UAVs and handheld sensors for ground operations. Avancerade Spaningssensorer is a three year project (2015-2017), where the activities in hyperspectral imaging during 2015 have been focused on sensor development and inventory and evaluation of software and algorithms for signal and image processing. An interferometric sensor has been developed in the long wave infrared spectral range, where spectral and radiometric calibration has been the main focus of attention during the past year. A new development project of a consumer camera, modified with a linear variable filter, has also been started. The sensor promises very high spatial resolution with the compactness and low weight of a normal system camera. In signal and image processing a MATLAB toolbox has been developed at FOI for hyperspectral imaging applications. The toolbox has been updated during 2015 with regards to both algorithm codes and manual. Evaluation has also been made of Opticks and SOCET GXP® for suitability as a graphical user interface for the toolbox as well as for creating plug-ins with hyperspectral functionality. During the next couple of years, the project will continue the activities in sensor development in order to provide demonstration capabilities of the technology to the Swedish Armed Forces. Studies will also be made in Deep Learning as a potentially new signal processing framework for hyperspectral imaging.