Research On Pre-stack Seismic Data Reflection Pattern Analysis Method

Seismic reflection mode analysis is an automatic seismic facies division by analyzing the seismic facies types of the seismic profile of the target horizon.Because of the high dimension and low signal-to-noise ratio of prestack seismic data,in order to obtain stable seismic facies analysis results,the current automatic seismic facies analysis technology mainly analyzes the post stack data,while the post stack seismic data,due to stack processing,hides the change rule of seismic wave with offset and azimuth in the pre-stack data,which reflects the change characteristics of underground rock strata more details.Based on the wide azimuth prestack three-dimentional seismic data,this paper studies the robust feature extraction algorithm of the data,analyzes and recognizes the seismic reflection mode,and uses the advantages of the pre-stack data information to improve the characterization ability of the features for different reflection modes,reduce the influence of the uncertain factors in the data on the mode analysis,so as to improve the accuracy and stability of the seismic reflection mode analysis.This paper first introduces the importance of seismic reflection mode in oil and gas exploration,explains the advantages of using pre stack seismic data for seismic reflection mode analysis compared with post stack data,and introduces the development of seismic reflection mode analysis method and related principles.To solve the problem of unsupervised seismic recognition based on pre stack seismic data,two feature extraction methods based on pre stack seismic data are proposed.The main work of this paper includes the following two aspects:(1)According to the characteristics of low signal-to-noise ratio of prestack seismic data,this paper introduces two-dimensional shearlet transform,and defines the shearlet feature of prestack seismic data by using the multi-scale and multi-directional characteristics of two-dimensional shearlet transform.In order to reduce the dimension of shearlet features,this paper proposes a new method of reducing the dimension of shearlet features by using the contraction convolution auto-encoder(CCAE).By introducing the feature penalty term into objective function of the convolution autoencoder(CAE),the influence of the uncertain factors in the seismic data on the reflection mode features can be suppressed as much as possible while the feature dimension is reduced.(2)As the existing auto-encoder network(AE)is based on one-dimensional and twodimensional data for dimensionality reduction,which is suitable for feature compression of poststack data or two-dimensional prestack seismic data,and can not use the spatial information of three-dimensional wide azimuth data,this paper introduces tensor model to express the wide azimuth data,and proposes a deep tensor auto-encoder network(TDAEN)based on t-product,which benefit from the expression ability of tensor.Compared with the traditional AE,the TDAEN has less network parameters and more sufficient feature expression ability.At the same time,for the TDAEN,a back propagation algorithm for iterative updating of network parameters is shown.In this paper,two feature extraction methods are applied to the reflection mode analysis of the synthetic data and the prestack data and wide azimuth data in the field.Through comparative analysis,the advantages of the proposed feature extraction method are verified.