Unsupervised Machine Learning: K-means Clustering Velocity Semblance Auto-Picking

Machine learning is becoming an attractive tool in various fields of earth sciences. During seismic data processing, velocity auto-picking can reduce time consumed on processing large volumes of seismic data and increase the number of velocity semblances which will be picked in a 3D seismic survey. In this paper, a new velocity auto-picking method on the base of unsupervised machine learning was proposed, a guide velocity as a constraint to pre-processing the velocity semblance was introduced and K-means clustering had been used to accomplish the velocity auto-picking. Real data processing shows that the proposed method is accurate and stable. The proposed method in the future will help geophysicist potentially reduce time spent on velocity semblance picking and obtain the relatively accurate time-velocity pairs effectively.