First Break Picking Based on Machine Learning

Summary Automatic or semi-automatic picking of the first break is an important task in the seismic data processing workflow, and incorrect picking can have rather serious impact on the quality of the processing results (Wong et al, 2021). In addition, manual picking of the first break is a long task (weeks or months) of routine monotonous work. This can introduce subjectivism into the processing procedure when the picking performed by different specialists can be also different. Also, dividing the territory into some segments for individual pickings adds additional sources for errors in the form of edge effects at the borders of these segments. All such dive errors can appear on speed maps as linear or segmental lanes. Automatic or semi-automatic picking of the first breaks can significantly reduce the level of the errors listed above and obtain uniform corrections over the area.