Application of deep learning in first break picking of seismic data
Yitao Pu, Xueli Zhang · 2018
With the application of broadband, wide-azimuth and highdensity (BWH) acquisition technique, the efficiency of acquisition has been dramatically improved. Meanwhile, the quantity of seismic data, especially for 3D seismic processing, has leapt from GB to TB, which sets a big challenge for both workers and machines. In statics correction, first break picking is the fundamental work and it decides the performance of statics correction. However, the method of first break picking still stays in traditional human-computer interaction currently, which is tedious and time consuming and cannot satisfy the request of large number of seismic data any more. The low efficiency may also delay the progress of projects. Convolutional Neural Network (CNN), one method of deep learning, has been used in large 3D seismic processing for the first time to solve the efficiency problem.