Machine‐guided velocity interpretation

In this paper we present a new approach for Some very innovative techniques, such as neural automatic interpretation of seismic stacking networks (Fish and Kusuma, 1994) and inversion velocities. The central characteristic of our (Landa et al, 1991) have been applied to this technique is that we derive velocity information problem. We believe that one of the reasons for initially at every time sample and at every CMP their lack of success, other than the apparent inertia location of the input data, and use automatic editing of the industry to change in this area, is that high to reject picks that are geologically inconsistent. computational costs force them to work with only a The machine-generated velocity field is then used small proportion of the information that is available. to overlay a conventional velocity analysis display Stacking velocities derived from CMP’s at, say, l/2 and guide the manual picking of the velocities, with km intervals, will have an inherent estimation error, the same options for QC and human intervention as regardless of how sophisticated the technique used in a conventional approach. to derive the velocities.