Artificial neural networks are highly simplified computer models that can simulate the human brain’s functionality to solve problems. A self-organizing map (SOM) is a type of neural network that uses a two-dimensional feature map to find significant patterns. In this paper, an SOM-based algorithm was designed to classify salt-contaminated stacking velocities and find the top salt surface simultaneously. The algorithm has been applied to the stacking velocities for a Gulf of Mexico (GOM) project. Results show that the SOM network could successfully classify the salt-contaminated velocities and with better resolution. The top salt surface classified by SOM has similar characteristics with the seismically interpreted model. Therefore, the SOM algorithm can be used to quickly derive a sediment velocity model to begin the depth migration process. In the future, SOM-based algorithms could be applied in preprocessing for the seismic pattern recognition to improve seismic interpretations.
Skip Nav Destination
Classification of Salt-contaminated Velocities With Self-organizing Map Neural Network
Paper presented at the 2001 SEG Annual Meeting, San Antonio, Texas, September 2001.
Paper Number: SEG-2001-0591
Published: September 09 2001
Zhang, Lin, Fortier, Al, and David C. Bartel. "Classification of Salt-contaminated Velocities With Self-organizing Map Neural Network." Paper presented at the 2001 SEG Annual Meeting, San Antonio, Texas, September 2001.
Download citation file:
Don't already have an account? Register