A 3D channel body interpretation via multiple attributes and supervoxel graph cut

Channels have always been vital geologic features in the exploration of hydrocarbon reservoirs, which makes the interpretation of channels an important task. Many different seismic attributes have been proposed to help the process of channel interpretation. A single seismic attribute could not fully and accurately reflect the geologic structure and edge details of a channel. Therefore, interpretation on a single attribute causes inaccurate segmentation. A 3D channel body interpretation method based on multiple attributes and supervoxel graph cut is applied in this paper, which identifies and segments the channel geologic body with fuzzy boundaries, poor continuity, or even data loss more accurately. First, a nonlinear dimensionality reduction method (locally linear embedding with geodesic distance) is applied to fuse a variety of seismic attributes to make channels clearer. Then, a graph-cut method based on the super geologic voxel is introduced, which reduces the computational complexity of segmentation and generates supervoxels more fitted to the edge of the channel body. Finally, a smooth 3D surface of the channel is obtained through the isosurface extraction. We use the data of a work area in northwest China and Parihaka-3D to evaluate the performance of our method. Our results show that, compared with other methods, the information provided by the fusion attribute is more complete, and the edge continuity of the channel is improved. The 3D channel bodies obtained by our method are clear and continuous. In the case of a complex channel body, our method can also work well.