A Pattern Recognition Approach for Automatic Horizon Picking in 3-D Seismic Data

Yu, Yingwei ¹; Kelley, Cliff ¹; Mardanova, Irina ¹
(1)Seismic Micro-Technology, Inc., Houston, TX.

We present a new approach to auto pick horizons in 3D seismic data using a pattern recognition technique.

This method was used as the basis of an innovative autotracking technology that has been utilized and proven by the interpretation community for almost two years.

In the practice of horizon picking, conventional methods usually employ a window-based approach in searching extrema. This approach only scans the adjacent trace vertically within a time window, while ignoring the lateral continuity. The very limited window context often incurs the “off cycle” effect, which means that the extrema points are incorrectly linked across seismic phase cycles. This effect can be more severe in seismic data with high-dip geologic structure. To preserve the lateral continuity of horizon picking, one needs to examine the seismic data pattern in a range of the neighborhood. The proposed method utilizes the context information to predict the horizon trend. We create a 3D data set constructed of directional pointers defined at every seismic data sample. It helps to determine the data pattern and direct the trace selection algorithm through the seismic volume. Combining the pointers set and a confidence-based trace selection mechanism helps to optimize the 3D horizon autopicking and obtain more accurate results compared to other conventional algorithms.

AAPG Search and Discovery Article #90135©2011 AAPG International Conference and Exhibition, Milan, Italy, 23-26 October 2011.