Classification for hyperspectral imaging

Adam Polak, Stephen Marshall, Jinchang Ren, David J.M. Stothard

Research output: Contribution to conferencePoster

28 Downloads (Pure)


Hyperspectral Imaging is a method of collecting and processing the information across pre-defined electromagnetic spectrum. These measurements make it possible to derive a continuous spectrum for each pixel of the image. After necessary adjustments these image spectra can be compared with database of reflectance spectra in order to recognise tested materials.
This project is conducted in cooperation between Fraunhofer Centre for Applied Photonics and Heriot-Watt Industrial Doctorate Centre in Photonics and Optics Technologies in partnership with University of Strathclyde. Fraunhofer Institute is known of world-class photonics solutions and this project aims in enhancement of one of their Hyperspectral Imaging systems with signal processing techniques. Set of classification procedures would be applied for the output of imaging spectrometer with the intention of spatial and spectral classification of objects captured by the spectrometer.
Spatial classification is based on Support Vector Machine (SVM) classifier. Use of texture features of the objects is considered as a base for labelling of detected items. Spectral classification is based on Partial Least Squares (PLS) method. With database of calibration reflectance spectra, method this can be used for prediction of “end members” concentration and therefore identification of the objects captured on the hyperspectral image. “
Original languageEnglish
Number of pages1
Publication statusUnpublished - 2014
EventIDC in Optics and Photonics Technologies Annual Conference - Heriot-Watt Univeristy, Edinburgh, United Kingdom
Duration: 17 Jul 201417 Jul 2014


ConferenceIDC in Optics and Photonics Technologies Annual Conference
Country/TerritoryUnited Kingdom


  • classification
  • Hyperspectral imaging
  • signal processing
  • spectral classification
  • partial least squares
  • spatial image classification


Dive into the research topics of 'Classification for hyperspectral imaging'. Together they form a unique fingerprint.

Cite this