Seismic facies recognition based on prestack data using two-dimensional Gabor transform and unsupervised clustering

In the field of oil and gas exploration, seismic facies maps are generally generated by the classification of prestack reflection structure information through pattern recognition technology, which plays an important role in determining underground oil and gas reservoirs. However, the prestack reflection structure signal has high dimensionality and complex information. It easily causes dimension disasters, large computational load, and inaccurate classification effect. Prestack reflection structure signal preprocessing, dimension reduction, feature extraction, and clustering algorithm are investigated, and a prestack reflection structure analysis method based on Gabor feature is proposed. The prestack reflection structure data are selected using reasonable isochronal interface as time window, and an enhanced Gabor feature is extracted by combining 2D Gabor transform. Then, the dimension of Gabor feature is reduced using principal component analysis method. Finally, unsupervised pattern recognition of dimension-reduced feature is carried out by combining with fuzzy C-means clustering algorithm. The practical application in the work area proves the superiority of Gabor feature in describing lateral variation and anisotropy of strata. Classification results based on Gabor feature can effectively distinguish different prestack seismic reflection structures and provide reliable basis for seismic facies analysis.