Seismic facies analysis is usually implemented using pattern recognition methods to classify seismic waveforms extracted along a horizon with a fixed time window. The commonly used similarity measures in those technologies are sensitive to the accuracy of horizon interpretation and the window selection and may thus lead to an unsatisfactory facies map. In this paper, we propose a novel similarity measure called dynamic subwindow matching for seismic facies analysis. The proposed measure is the average similarity of a series of short waveforms containing optimal matching subwindows instead of using the whole signal as done in traditional measures. Because of the introduction of nonlinear alignments for seismic reflection events, the proposed measure has the ability to overcome the horizon inaccuracy problem in seismic facies analysis. Furthermore, it can be directly applied to measure the similarity of the waveforms with different time lengths from the targeted interval. Finally, a new method combining the proposed measure and the k-nearest neighbours method is developed for supervised seismic facies analysis. The synthetic and field data application demonstrate that the proposed measure is robust to amplitude and horizon interpretation noise and provide an effective approach for seismic facies analysis of varying-thickness reservoirs.

No abstract