CEES-SEP
Students at the Stanford Exploration Project SEP have been working closely with the Center for Computational Earth and Environmental Science CEES.
This collaborative effort brings the state-of-the-art computer resources of CEES to bear on petroleum-related geophysics problems at SEP. Leading this effort for SEP is Bob Clapp.
PROJECTS:
Lomask, J., Clapp, R. G., and Biondi, B., submitted 2006, Image segmentation for tracking 3D salt boundaries: Geophysics.
We use a modied version of the normalized cuts image segmentation (NCIS) method to partition seismic images along salt boundaries. We identified two steps of this algorithm that could benefit from parallel implementation. The first step is the calculation of the weight matrix, where we can use a parallel cluster to signicantly increase computation speed. The second is the estimation of the eigenvector with the second smallest eigenvalue, where parallel clusters can hold the entire sparse matrix in memory rather than storing and retrieving sections from disk. The tracked interface (red) is generated from the segmentation algorithm and compares well to a manually picked result (yellow).
|
