Seismic First Arrival Picking Using Grey-Level Co-Occurrence Matrix-Based Texture Analysis

A novel technique for automatic seismic first arrival picking is presented. Our method relies on the grey-level co-occurrence matrix to detect seismic time samples that share common displacement and orientation features, enabling us to separate uncorrelated events while preserving correlated seismic events. For each seismic time sample, four distinct grey-level co-occurrence features are extracted: energy, contrast, homogeneity and correlation. Based on these, each time sample is then classified using fuzzy C-means into either the noise segment or the correlated events cluster. The correlated events segment is subsequently examined to identify the precise first arrival picks by selecting the earliest samples whose short-time average to long-time average ratio values exceed the mean. Tests on real seismic shot records demonstrate the superiority of our proposed method in accurately picking the first arrivals, with an average improvement of more than 14% in picking precision, while producing the lowest histogram error. Our suggested method also shows strong performance in recognizing first arrival picks in poorly acquired (scattered) and faded traces, where other reported methods often fail.