The neural net of Carpenter/Grossberg's adaptive resonance theory (ART2) has the ability of the unsupervised self-organizing clustering for analog input patterns. So we use the ART2 neural net for seismic wavelet clustering in the seismogram. The preprocessing includes peak detection and wavelet extraction using the supervised multilayer perceptron (MLP). After extracting seismic wavelets, we use ART2 net to cluster the seismic wavelets. In the experiment, the real seismogram at Mississippi Canyon is analyzed. 371 wavelets are extracted and clustered into 3 classes. The wavelet clustering results can show the uniform property of the reflection layer and improve the seismic interpretation.
|Number of pages||4|
|Journal||SEG Technical Program Expanded Abstracts|
|State||Published - 1 Jan 2000|
|Event||2000 Society of Exploration Geophysicists Annual Meeting, SEG 2000 - Calgary, Canada|
Duration: 6 Aug 2000 → 11 Aug 2000