Disclosed herein is a method for recursive hierarchical data segmentation which eliminates the processing window artifacts that often occurred in a previous recurisve approach. This new approach is compared to previous recursive approaches and to a non-recursive coarse-to-fine processing approach that avoids the creation of the processing window artifacts altogether. Parallel implementations of the new recursive approach and the contrasting coarse-to-fine approach are discussed. The parallel implementation of the recursive approach utilizes a variant of the patent pending implementation approach originally disclosed in NASA Case No. 14,305. The recursive version has a much more efficient parallel implementation, making it the best choice for processing large data sets.Currently being used in the following NASA projects:"Knowledge Discovery and Data Mining Based on Hierarchical Segmentation of Image Data," funded 2001-2004 by NASAs Intelligent Systems Program under NRA2-37143. Principal Investigator: James C. Tilton, NASA GSFC. Greenbelt, MD. Co-Investigators: Giovanni Marchisio, Insightful Corp., Seattle, WA; and Mihai Datcu, German Aerospace Center, DLR, Oberpfaffenhofen, Germany."Applying Intelligent Systems Technology to Extract and Understand Radio Imaging Data: An Application to Radio Plasma Imager Data from the IMAGE Mission," funded 2003-2005 by NASAs Intelligent Systems Program as a Mission Infusion Task. Principal Investigator: Shing F. Fung, NASA GSFC, Greenbelt, MD. Co-Investigators: Scott A. Boardsen, L3 Communications Analytics Division, Lanham, MD; Leonard Garcia, QSS Group, Inc., Lanham, MD; Robert F. Benson and Vladimir Lumelsky, NASA GSFC, Greenbelt, MD; Ivan Galkin and Bodo Reinisch, Center for Atmospheric Research, University of Massachusetts, Lowell, MA; and James C. Tilton, NASA GSFC, Greenbelt, MD (IS Technology Partner).In the "Knowledge Discovery and Data Mining" projects, HSEG/RHSEG is being used to generate hierarchical image segmentations that are being used as the basis for indexing image information for use in VisiMine, a knowledge discovery and data mining system developed by Insightful Corporation. The HSEGViewer program is being used to prototype techniques for quantifying the information content conveyed by the hierarchical segmentations. HSEG/RHSEG and HSEGViewer are the key elements, along with VisiMine, of this research project. Earth science imagery databases are targeted for knowledge discovery and data mining in these projects.In the "ApplyingtoRadio Imaging Data" project, HSEG/RHSEG is being used to generate hierarchical image segmentations that are being investigated for their utility in identifying features of interest in the Radio Imaging Data. The HSEGViewer program is being used to prototype techniques for automatically identifying features of interest. HSEG/RHSEG and other software to be developed based on HSEGViewer prototyping is planned for use in enhancing the existing Radio Plasma Imager Feature Identification Software, CORPRAL (Cognitive Online RPI Plasmagram Ranking Algorithm).
Publications
Not Available
Technical Support Package:
Not Available
Related TOPS
Patent Number:
Not Available - Patent Pending
Licensing:
Not Available
Public Release Date:
9/19/2006
Reference Number: GSC-14681-1
|