The multilayer perceptron neural network is trained as a classifier and is applied to the recognition of seismic patterns. The principle of training the multilayer perceptron is described. Three classes of seismic patterns are analyzed in the experiment: bright spot, pinch-out, and horizontal reflection patterns. Seven moments that are invariant to translation, rotation, and scale, are employed for feature generation of each seismic pattern. The training set includes noise-free, low-noise, and misclassified seismic patterns. The test set includes seismic patterns with various noise levels. The multilayer perceptron is initially trained with the training set of noise-free and low-noise seismic patterns. After convergence of the training, the network is applied to the classification of the test set of noisy seismic patterns. Some misclassified patterns with higher noise level are added to the training set for retraining. The classification and training process is repeated through several stages. This retraining can significantly improve the robustness of the network. The converged network at each training stage is applied to the real seismic data at Mississippi Canyon, the bright spot pattern can be detected after the retraining at the higher noise level. From these experiments, the multilayer perceptron is shown to have the capability of robust recognition of seismic patterns.
|Number of pages||6|
|State||Published - 1 Jan 2001|
|Event||International Joint Conference on Neural Networks (IJCNN'01) - Washington, DC, United States|
Duration: 15 Jul 2001 → 19 Jul 2001
|Conference||International Joint Conference on Neural Networks (IJCNN'01)|
|Period||15/07/01 → 19/07/01|