基于广义Curvelet变换和相空间特征聚类的VSP波场分离方法

垂直地震剖面(Vertical Seismic Profiling, VSP)蕴含大量的地层地质信息, 是连接地震反射数据与测井资料之间的桥梁.VSP数据分辨率高, 资料丰富, 包括上行波、下行波、转换波等多种类型的波场, 常被用于地层反射界面标定、介质衰减参数反演等.其中, 如何有效地分离波场是将VSP资料应用于油气勘探的关键之一.广义Curvelet变换是一种具有多尺度、多角度的相空间变换, 能够进行相空间域波场特性的提取.本文聚焦于VSP波场的上下行波分离问题, 提出一种基于广义Curvelet变换和相空间特征聚类方法.首先, 采用广义Curvelet变换对VSP波场进行特征提取, 构造融入相空间角度信息的特征数据集; 然后, 利用K-means聚类算法在相空间对波场特征进行分类; 最后, 对分类结果进行反变换, 完成VSP上下行波波场的自适应分离.为了验证方法的有效性, 本文将所提出的方法用于合成数据和实际VSP数据的波场分离, 并与常用的基于F-K变换的波场分离方法进行对比.处理结果表明, 本文方法角度分辨率高、抗噪能力强, 波场分离结果的保真保幅性好, 这为后续的成像与地层的特征分析提供了重要基础资料.

VSP wavefield separation based on GCT and clustering algorithm combining with phase space features

Vertical Seismic Profiling (VSP) contains a large amount of stratum geological information and is a bridge between reflection seismic data and well logging data. VSP data is high-resolution and rich in information, which contains upgoing waves, downgoing waves, transition waves and other types of wavefields. The VSP data is always used for the calibration of the geological reflection interface and the inversion of the medium attenuation parameters. How to effectively separate wavefield is one of the keys to applying VSP data in oil and gas exploration. The Generalized Curvelet Transform(GCT) is a multi-scale and multi-directional phase space transform, which can extract characteristics of the wavefield in phase space. In this paper, we focus on the upgoing and downgoing waves separation of VSP wavefield and propose a VSP wavefield separation method based on GCT and phase space feature clustering. Firstly, it uses the GCT to extract features from the VSP wavefield and constructs a feature data set incorporating phase space angular information. Further, the K-means clustering algorithm is used to classify the wavefield features in the phase space. Finally, the inverse transform is performed to the classifications to achieve the adaptive separation of the upgoing and downgoing waves of the VSP data. In order to verify the effectiveness of the method, the proposed method is applied to the wavefield separation of the synthetic data and field VSP data, and compared with the commonly used F-K transform-based wavefield separation method. The results show that the proposed method has high angular resolution, strong anti-noise ability, and well fidelity and amplitude preserving, which can provide important basic data for the subsequent imaging and the analysis of stratigraphic characterizations.