AISRP workshop '97 presentation,    E. Merényi, U of Arizona, LPL

4. Focus of this Project

To produce a parallel  software package for clustering/classification of hyperspectral signatures, ready for parallel hardware implementation at the end of the grant period. note 4.1

• Research/improve  ANNs and other related algorithms for high-D data and many classes

• Engineer/integrate ANN classifiers with visualization and evaluation modules (existing prototypes, plus new functions)
• Platform independency, friendly GUI and help
• Concept proving with real space /airborne data (Mars Pathfinder, Clementine, AVIRIS, ...)

4.1  Realization of Goals

Algorithm developments

• are based on improving and integrating existing successful prototypes;
• utilize off-the-shelf tools :
  • NeuralWorks Professional II/Plus, Designer Pack and User Written NeuroDynamics by NeuralWare, Inc., for  Artificial Neural Net design, testing and deployment in C code; these packages are proprietary but deployed networks are not;
  • Tcl/Tk for GUI development -- free
  • Khoros (Khoral Research,  Inc., University of New Mexico) for visualization and image manipulation -- free
  • C/C++, XWindows, standard plotting packages, UNIX shell scripting

• are tested with well understood and already analyzed large spectral images of Earth (AVIRIS), the Moon (Clementine), and Mars (telescopic data). The new Mars Pathfinder images are the exception, providing possible unknown challenges to respond to.