A First-Arrival Picking Technique Based on Texture Segmentation Exploring Seismic Data

In this study, we propose an innovative first arrival picking technique based on texture segmentation of seismic shot records for exploring seismic data. The seismic shot records are divided into clusters (depending on the seismic event types) based on texture-extracted features and fuzzy c-means. This technique utilizes the industrial energy ratio procedure to be conducted prior to clustering and applied to the cluster containing the first arrivals to recognize seismic shot points corresponding to the direct arrival picks. The suggested procedure was tested on one synthetic and two real seismic shot records. Using the proposed technique, a pick accuracy of more than 99% was achieved for the synthetic data set with a noise level of 10%, and more than 80% accuracy was achieved for the real data shot records, and all tests were within an absolute error tolerance of ±20 ms. Additionally, the proposed technique picks were more accurate than the picks of the standard industrial Coppen's method as well as the projection onto convex sets segmentation technique by an overall average accuracy of approximately 28.98%.