First arrival picking is the basis of the following seismic processing and imaging. The most widely used method, manual picking, is labor intensive, time-consuming as the seismic data volumes grows sharply. Robust and automatic picking methods need to be developed to meet the challenges. We propose a novel method to automatically pick the first breaks using an improved 2D pixel-wise convolutional network. Events picking is transferred into a problem of binary image segmentation through respectively tagging signal before and after first arrivals as one and zero. Field examples with borehole seismic data are used to demonstrate the effectiveness in reduce labor and superiority over traditional automatic picking methods.
Presentation Date: Monday, September 16, 2019
Session Start Time: 1:50 PM
Presentation Time: 3:55 PM
Location: Poster Station 2
Presentation Type: Poster