Proposing new seismic texture attributes based on novel gray level matrix with application to salt dome detection

Abstract

The exploration of new hydrocarbon resources requires a detailed image of the subsurface geological structures. Interpreting seismic sections is one of the most common ways of accurately imaging the Earth's subsurface. Automated seismic section interpretation requires accurate delineation of the target geobody through seismic section segmentation. Texture analysis of images is one of the common tools for seismic section segmentation for target geobody identification. Exploration of geological phenomena e.g. the salt dome, buried channels, etc., is very important in the field of hydrocarbon exploration and production due to the possibility of creating stratigraphic and structural hydrocarbon traps, creating potential for subsurface energy storage and drilling hazards. They have textural differences with their surrounding environment, and therefore the analysis of seismic sections using textural attributes to determine the geometry of these events is one of the challenges facing interpreters. Gray level co-occurrence matrix (GLCM) is the commonly used tool for textural analysis of seismic images, based on which several attributes have been introduced. Optimal adjustment of numerous input parameters in GLCM attributes and their strong dependence on the dip of events are the drawbacks of this method. In this paper, we proposed new seismic attributes based on the newly introduced gray level matrices (GLM) consisting of gray level run length matrix (GLRLM), gray level size zone matrix (GLSZM), gray level difference matrix (GLDM), and normalized gray level dependence matrix (NGLDM). The new proposed seismic attributes depend on fewer input parameters for adjustment than conventional attributes while increasing accuracy in event detection, and even GLSZM-based attributes are independent of the phenomena dip. The efficiency of the proposed attributes was evaluated on the real field and synthetic seismic data containing a salt dome and its results were compared with conventional GLCM-based attributes. The qualitative and quantitative results showed that in addition to the methodological superiority of the newly introduced gray level matrices compared to the GLCM, the accuracy of the proposed attributes was also increased in the salt dome detection. Moreover, it seems that the linear transform to grayscale performed better than the non-linear one in distinguishing the salt dome from the surrounding sediments. But the main challenge is distinguishing the salt dome texture from the weak layering which the nonlinear transform has done better than the linear one.