Neural networks for robust recognition of seismic reflection patterns

In this paper, a neural network approach for the recognition of seismic reflection patterns is developed. A multi-layer perceptron is trained as the classifier by using the back-propagation algorithm which varies the number of hidden units [1], and the central moments are employed for feature extraction. Seismic reflection patterns (classes) are designed in the experimental domain. The network is initially trained with noisefree training samples, and is retrained gradually with misclassified noisy shifted testing patterns to improve the robustness of the classifier. Through classifying a large set of unknown testing patterns of various noise degrees, a great augmentation in system robustness and an encouraging recognition performance are presented.