Processing Radio Plasma Imager Plasmagrams Utilizing Hierarchical Segmentation    

I.A. Galkin, G. Khmyrov, J.C. Tilton, S.F. Fung, and B.W. Reinisch   


The Radio Plasma Imager (RPI) on the IMAGE spacecraft provides valuable
information on the remote and local electron densities in the Earth's
magnetosphere. This information is derived from the RPI plasmagrams, a common
visual representation of active radio-sounding data in the format of echo
intensity as a function of echo delay time (ordinate) and sounding frequency
(abscissa). Due to the large volume of the archived RPI plasmagram imagery,
automated data exploration software, ÒCognitive Online Rpi Plasmagram Ranking 
AlgorithmÒ (CORPRAL), has been developed to identify plasmagrams containing
signatures of interest. CORPRAL routinely scans the RPI mission database to
select qualifying plasmagrams. The echo detection algorithms implemented in the
 CORPRAL, still yield significant "false-alarms" and thus can benefit from the
assistance of additional image processing techniques. As a NASA CICT/IDU
Technology Infusion task, we are exploring the adaptation and incorporation of
the recursive hierarchical segmentation (RHSEG) technique that was developed
for a broad class of remotely sensed images of the Earth. The RHSEG algorithm
iteratively builds a hierarchy of image segmentations of various detail levels
using a hybridization of region growing, and constrained spectral clustering.
Analysis of the segmentation hierarchy allows one to track the characteristics
of each region in the process of its growing to optimally select the best
segmentation level to describe the corresponding feature. Treatment of
plasmagrams with the RHSEG will then provides good candidate signatures for 
registration, thus improving the robustness of CORPRAL. This paper discusses
our progress to date. 

_______________
Proceedings NASA Earth Science Technology Conference (ESTC), B8P3: 1-7, 
June 22-24, 2004