Unsupervised Seismic Facies Pattern Recognition Based On 3D Gabor Transform And Two-Step Cluster Analysis
Conventional methods usually use poststack or prestack data to divide seismic facies directly, ignoring the spatial structure characteristics of 3D seismic data. In this study, a seismic facies statistical pattern recognition method based on 3D Gabor transform and two-step cluster analysis is proposed to address the weak discrimination and poor continuity of conventional seismic facies recognition techniques. The feature extraction of the seismic reflection signal is performed by using a multiscale and multidirectional 3D Gabor filter, and the unsupervised statistical pattern recognition of seismic facies is executed through a two-step cluster analysis. The results of the seismic simulation data testing and its application in actual working areas show that the proposed method not only improves the identification accuracy of fault and fracture, but also significantly enhances the spatial continuity of the seismic facies map. The 3D Gabor transform can utilize the information of the spatial structure contained in the seismic data. Combined with two-step cluster analysis, the proposed approach can effectively address the problems regarding the excessive amount of data processing and difficult classification of discrete values .