Abstract
Geological structural and target stratigraphic units can be well distinguished and clearly presented in seismic data based on their physical characteristics. However, due to the inherent band limited and noise contaminated and sometime distorted image on seismic data, it is not straight forward to identify, delineate and image those targets in seismic data. In this regard, the seismic attributes are used as a tool for extracting valuable information and characteristics from the seismic data for further appropriate geological interpretation of the target. In large size 3D seismic data, extreme interaction time with costly labor force is required for the manual interpretation of the seismic images. Automation of this procedure besides intelligent analysis of large number of various combination of attributes with different computational algorithms provides a reliable, fast and high productive alternative path. In this study, an automatic object delineation method based on combination of seismic attributes is developed based on the graph theory, mainly to identify the salt dome boundary in seismic data. Results are compared with the MLP and weight of evidence methods. For this purpose, initially a set of attributes with different window lengths are extracted from the data and then by using the principal component analysis method, the most appropriate set of attributes for combining are determined. Then a corresponding graph is derived for each attribute and the distance matrix of each graph is calculated and finally the similarity matrix is defined. Subsequently, the graphs are combined using the distance method. The binary model then is obtained by applying a threshold and image segmentation technique. To evaluate the results, they are compared with the interpreted models from the leading-edge methods, and for quantitative comparison, the accuracy value of the method then is calculated pixel by pixel. Based on the quantitative comparison of results, the accuracy obtained by the proposed method here is acceptable and is higher than the well-known MLP and weight of evidence methods, which shows the efficient application of the graph-based combination method in combining seismic attributes and delineation the salt dome boundary.
NSTL