Abstract:

We present a method for automatic events extraction based on fourth-order cumulant in the case of that events can’t be identified under the condition of powerful stochastic noise in seismic exploration. This method makes use of fourthorder cumulant to estimate the time delay, and to avoid correlative Gaussian noise. And in the selection of reference trace, the principle of setting multiple reference trace in certain interval reduced the cumulate error of events extraction. The theoretical model computation result shows that this method has more powerful ability to restrain the correlative Gaussian noise than cross-correlation method, the signal-to-noise ratio reaches -5 dB. In the implementation of real data, the parameters related to events such as events position and formation dip had extracted. The parameter provides a basis for attribute analysis and quantitative interpretation of this seismic data.

Key words: fourth-order cumulants, time delay, automatic extraction, signal to noise ratio, correlated colored Gaussian noise