Research On Seismic Interpretation Algorithms

Seismic interpretation which transforms seismic information into geological in-formation is an import part of seismic exploration. The seismic interpreters analyze the seismic data to get the information about geological structures. The results of seismic interpretation will make a great impact on seismic exploration. Over the past few decades, the seismic interpretation techniques have got rapid development. With the development of seismic prospecting, the seismic datasets become more and more huge. Moreover, the seismic interpretation techniques become more and more com-plicated. How to improve efficiency and accuracy of seismic interpretation processes is a research hotspot in the field of seismic interpretation.Two fundamental seismic interpretation techniques are studied in this paper. One is seismic coherency cube, the other is horizon automatic picking. Coherency cube is an important seismic attribute and it can be used to detect some irregular geological structures (e.g., faults). Horizons are surfaces which separate different underground rock layers. Horizon picking is an essential process to some other seismic interpreta-tion processes. In this paper, we try to improve the efficiency and accuracy of the two processes, using GPU general-purpose computing techniques and image processing techniques. The main work of this paper is summarized as follows:(1) A CUD A based C1coherency cube algorithm is proposed. Coherency cube is a time-consuming process. As Cl coherency cube algorithm is a data-parallel compu-tation, it is very appropriate to be implemented in GPU. We implement C1coherency cube algorithm using CUDA. Experiments show that the algorithm has a significant speedup comparing with the CPU implementation of C1coherency cube algorithm.(2) A CUDA based horizon tracking algorithm is proposed. The traditional hori-zon auto-picking algorithm is a seeds-based growing (or tracking) process and it is an iterative algorithm which is not easy to be implemented using GPU. We extract the data-parallel parts of the horizon auto-picking and implement them using CUDA. The effectiveness of our algorithm is verified by experiments.(3) An energy minimization based horizon picking algorithm is proposed. We introduce an energy function which reflects the seismic image features. The horizon picking problem is transformed into an energy minimization problem by using the energy function. Then the energy minimization problem is solved by graph cuts which are widely used for image segmentation. Experiment results have verified the effi-ciency of our algorithm.