Advancing Seismic Image Segmentation: UNET++ with GLCM Integration

Abstract

Seismic image segmentation is a critical task in geophysical exploration, facilitating the identification of subsurface geological structures essential for resource assessment and risk mitigation. Traditional manual segmentation methods are laborious and subjective, highlighting the need for automated techniques to enhance efficiency and accuracy. Leveraging the advancements in deep learning, this study proposes a novel methodology for seismic image segmentation by integrating the UNET++ architecture with Gray-Level Co-occurrence Matrix (GLCM) features. This approach aims to achieve higher segmentation accuracy, reduce processing time, and improve generalization capabilities. The methodology is validated using the TSG Salt dataset, and extensive experimentation demonstrates its superior performance compared to existing approaches. Results indicate significant enhancements in segmentation accuracy and computational efficiency, positioning the proposed methodology as a promising advancement in seismic imaging techniques for geological analysis and resource exploration.