We consider the task of channel detection as a regression problem and propose a workflow to identify the complex channels in 3D seismic volumes with complex structure using an end-to-end 3D convolutional neural network. To train the network, we automatically generate a training dataset containing more than 19000 3D synthetic seismic volumes and the corresponding channel labels, which are shown to be sufficient to train a good channel identification network. After training with only the synthetic dataset, the network automatically learns useful features that are important for channel identification. Multiple synthetic and field examples show that the network can much more accurately and efficiently predict channels in 3D seismic volumes with complex structure.

This content is only available via PDF.
You can access this article if you purchase or spend a download.