Automatic velocity analysis with physics-constrained optimal surface picking

Abstract

Stacking velocity is generally obtained by picking the energy peaks in velocity semblance which can be time consuming when performed manually. Numerous automatic methods have been proposed for accelerating the velocity picking but often generate physically unreasonable picking results when strong and spatially consistent energy anomalies (due to noise and multiples) appear in the semblance map. We develop a constrained optimal surface picking method to automatically pick a 2D velocity field from a 3D semblance volume with high efficiency and robustness. This method is improved from the 2D dynamic programming algorithm by incorporating vertical physical constraints in the time direction and lateral smoothness constraints in the common-midpoint (CMP) direction. The time-direction physical constraint ensures that the picked velocity is positive when converted to an interval velocity, whereas the CMP-direction smoothness constraint ensures that the picked 2D velocity field is laterally continuous. Tests on the Marmousi-2 model indicate that our constrained optimal surface picking algorithm improves the spatial structure consistency of the picked velocity field and is able to robustly avoid picking the strong and consistent energy features generated in the semblance volume by multiples. We further determine the robustness of our method in a 2D real data example by comparing our automatic picking from a 3D semblance volume with a result that is manually picked from individual 2D semblance slices. The comparison indicates that the general trend of our result is consistent with the manual picking and our result looks geologically more reasonable in detail and generates a better stacking image with improved, focused, and more continuous reflections.