Statistics
Scientific paper
Dec 2004
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2004agufm.p43a0912b&link_type=abstract
American Geophysical Union, Fall Meeting 2004, abstract #P43A-0912
Statistics
5470 Surface Materials And Properties, 5494 Instruments And Techniques
Scientific paper
We have developed the Rock Identification Toolkit (RockIT), a cross-platform, graphical program designed to help geologists rapidly and accurately label rocks in images and report both per rock and per image statistics. We divide rock labeling into two major phases. First, RockIT calculates initial traces (boundaries) for rocks in a scene by arbitrating (combining) Sobel and Canny edge detectors at multiple scales. While this phase is quick, at seconds to minutes per image, it may be performed in a non-interactive, batch mode, so as to decrease the total time spent in front of RockIT. The initial machine-assisted rock determination is not perfect and becomes less so for certain classes of images (e.g. scenes with high dust deposition or images with low contrast). In the second phase, a person may improve upon inaccurate rock detections. For this, RockIT provides a set of tools, like Polygon, Pencil and Zoom, inspired by well-known paint program metaphors, to allow modification, deletion or addition of rock traces. Finally, to further speed labeling, RockIT provides a Snake tool to wrap rough-cut traces tightly around rock perimeters. As rock traces are added, removed and modified, RockIT continually updates a table of per rock statistics. The most basic statistic reports rock area (in pixels). A more involved set of statistics uses a direct least squares technique to fit a rock trace to an ellipse and report the semimajor and semiminor axes, orientation, eccentricity and quality of fit. In the future we plan to add additional statistics to characterize rock albedo and texture. JPL geologists have used RockIT to label and analyze both the microscopic imager and panoramic scenes from the Mars Exploration Rovers mission.
Andreson R. C.
Bornstein Benjamin
Castano A.
Castano Rebecca
No associations
LandOfFree
RockIT: A Graphical Program for Labeling and Analyzing Rock S cenes does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with RockIT: A Graphical Program for Labeling and Analyzing Rock S cenes, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and RockIT: A Graphical Program for Labeling and Analyzing Rock S cenes will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1453812