基于CLG光学流和波场分解的逆时偏移角度道集提取方法研究

RTM angle gathers based on the Combining Local and Global(CLG)optical flow method and wavefield decomposition method

角度域共成像点道集是衔接叠前地震数据与储层特征的重要桥梁,对地震偏移成像与储层描述具有重要意义.与克希霍夫偏移和单向波动方程偏移相比,逆时偏移是复杂地区最精确的成像方法.高效稳健地生成逆时偏移角度道集目前仍然是一个挑战.本文主要讨论如何采用光学流方法高效、高质量地提取角度道集.在逆时偏移波场外推过程中,光学流方法可以估计波场传播方向.其中Lucas-Kanade(LK)和Horn-Schunck(HS)方法是光学流方法中两种典型的方法.LK光学流方法是一种局部方法,该方法依赖于局部点的梯度值,但是容易出现奇异现象,HS光学流方法属于全局方法,波场方向估计依赖于整个波场,易受噪声影响,对异常值比较敏感,导致整体波场方向计算精度不高.本文提出采用局部和整体结合(Combining Local and Global,CLG)的光学流方法估计波场传播方向.该方法可以有效地提高波场方向的精度,并且简单高效,便于并行处理.对比HS光学流方法,CLG光学流方法几乎不增加额外的计算量.另外,为了弱化光学流方法无法处理波前重叠问题,本文利用解析波场和方向滤波对波场进行方向分解,仅需波场的空间傅里叶变换即可实现任意波场方向分解,将分解后的波场分别估计波场反向,提取成像结果.进一步地,在估计反射张角和方位角时,本文提出有效的归一化方法和改进的最小二乘除法,提高角度估计的精度和稳定性.最后,理论和实际资料例证了本文提出方法的有效性.

Angle-Domain Common-Image Gathers(ADCIGs)is an important bridge connecting the pre-stack seismic data and reservoir characteristics,which is of great significance for seismic migration and reservoir characterization.Compared with Kirchhoff migration and one-way wave-equation migration,Reverse-Time Migration(RTM)is the most accurate imaging method for complex areas.Generating the RTM angle gathers with efficiency and robustness still is a challenge.In the paper,we will mainly discuss how to efficiently generate high-quality angle gathers with the optical flow method.The optical flow method can be applied for the wavefield direction estimation,during the process of RTM wavefield extrapolation.The Lucas-Kanade(LK)technique and the Horn-Schunck(HS)algorithm are two fundamental approaches.However,the LK method relies on local wavefield gradient and suffers from the singular phenomenon.The HS method is sensitive to noise,and the accuracy of the estimated wavefield direction is not high.We propose using the Combining Local and Global(CLG)optical flow method to estimate the wavefield direction.The accuracy of wavefield direction can be improved effectively by the proposed CLG optical flow method,especially for the reflected interface.And the proposed method is simple,efficient and convenient for parallel processing.Moreover,compared with the HS optical flow method,the proposed method has not much extra calculation.Then,we apply the wavefield decomposition method with the analytical wavefield and directional filtering to weaken the inaccurate estimation of optical flow method at the wave-front overlapping situation.In the analytical wavefield,the directional decomposition can be realized by just only the spatial Fourier transform of the wavefield.Then we calculate wavefield direction and extract angle gathers based on the decomposed wavefield.Furthermore,an effective normalization method and an improved least square division method are proposed to improve the accuracy and stability of angle estimation.Finally,theoretical and practical data illustrate the effectiveness of the proposed method.