Abstract

Seismic fault interpretation is particularly vital for localizing hydrocarbon reservoirs. Beyond the mere detection of probable fault points, extracting discrete fault surfaces has emerged as a critical yet underexplored task in subsurface geological analysis. This paper mainly focuses on developing an automated fault tracking framework. This framework begins with fault line enhancement through morphological operations to refine features and reduce noise. Subsequently, it employs the Probabilistic Hough Transform for robust fault line detection. Furthermore, we utilize an unsupervised density-based clustering algorithm (DBSCAN) to effectively group these detected line segments into coherent fault surfaces, addressing the varying densities inherent in geological fault distributions. The paper also presents comprehensive experimental results on both synthetic and real-world Equinor model datasets.