Nonlinear Set to Set Pattern Recognition

Applications in Hyperspectral Imaging and More

At a Glance

Researchers at Colorado State University have developed a novel pattern recognition system, wherein variations in the states of patterns can be exploited for their discriminatory information. The system has the ability to compare data sets of unlabeled patterns (having variations of state) in a set-by-set comparison with labeled arrays of data sets of multiple patterns (also having variations of state). Ultimately, the system has the ability to classify and/or identify images – even in the case of hyperspectral imaging.


A hyperspectral image is a data cube with two spatial dimensions and one spectral dimension. Typically, initial data set of images tends to include substantial variations of state (e.g., in illumination, position, etc.) that make the evaluation with the unknown image difficult to resolve. In one existing approach, illumination and/or position variations, for example, in the data set of images are removed by computing illumination and/or invariant images to obtain a more normalized data set. However, such normalization obfuscates unique characteristics leading to improper recognition and/or inability to identify such images.


  • Ability to create separable data sets in hyperspectral imaging
  • Natural for representing nonlinear spaces
  • Can be implemented in parallel
  • Reliably identification from data sets with large variations
  • Resolution requirements are minimal (commercially viable computing requirements)


  • Hyperspectral imaging
  • Precision medicine (e.g., Cancer detection, diagnostics, etc.)
  • Facial recognition in uncontrolled environments (i.e. not an artificial studio)
  • Security and video surveillance (e.g., used in banks, airports, govt. buildings, retail stores)
Last Updated: April 2022
Hyperspectral Imaging Visual

Available for Licensing

IP Status

Michael Kirby
Christopher Peterson

Reference Number
Licensing Manager

Jessy McGowan