A New Data-Driven Seismic Interpretation Workflow Using Unsupervised Machine Learning and Non-Local Trace Matching
A.J. Bugge, J.E. Lie, A.K. Evensen and S. Clark
Event name: 81st EAGE Conference and Exhibition 2019
Session: AI/Digitalization for Interpretation - Geological and Model Building Interpretation
Publication date: 03 June 2019
Info: Extended abstract, PDF ( 1.02Mb )
Price: € 20
Extracting key horizons from seismic data is an important element of the seismic interpretation workflow. Numerous computer-assisted horizon extraction methods exist, but they are typically sensitive to structural and stratigraphic discontinuities and have difficulties to extract non-coherent dislocated horizons. We propose a new data-driven seismic interpretation workflow to extract dislocated seismic horizons from complex 3D seismic data. The proposed workflow combines two separate methods: identification of stratigraphic sequences using unsupervised machine learning, and automatic correlation and extraction of seismic horizons using non-local trace matching. We use unsupervised machine learning to automatically identify and label stratigraphic sequences, and non-local dynamic time warping to iteratively track and extract seismic horizons from each of the identified sequences, constrained by the sequence boundaries. This workflow allows us to interpret structurally and stratigraphically complex seismic volumes completely automatically, and it require no interpretive experience or geological knowledge.