叠前地震数据特征波场分解、偏移成像与层析反演

本文提出了一套叠前地震数据稀疏表达(特征波场合成)、深度偏移成像和层析成像的处理流程.不同于传统的变换域中的数据稀疏表达理论,本文利用局部平面波的传播方向(慢度矢量),在中心炮检点处同时进行波束合成,从而将地震数据投影到局部平面波域(高维空间)中.由于波束合成后的地震数据描述了局部平面波的方向特征,因此称之为特征波场.然而波束合成算法需要估计局部平面波的慢度矢量.当地震数据受噪声干扰时,难以在常规τ-p谱中自动估计局部平面波的射线参数(慢度矢量).本文提出了基于反演理论的特征波场合成方法,可以同时反演局部平面波及其传播方向,从而提高特征波合成的自动化程度并保持方法的稳健性.通过特征波场合成,可以将地震数据分解为单独的震相(波形).这样的数据可以直接用来成像及反演.在局部平面波域中,由于局部平面波的入射与出射射线参数已知,传统的Kirchhoff叠前深度偏移(PSDM)和高斯束/控制束PSDM可以实现从"沿等时面的画弧"到"向反射点(段)的直接投影"的转变,叠前偏移的效率以及成像质量可以同时提高.此外,特征波场与地下反射点(段)的一对一映射关系使得叠前深度偏移与层析成像融为一体,可以极大地提高速度反演的效率.数值试验证明了特征波场合成、叠前深度成像以及层析反演的有效性.

Characteristic wavefield decomposition,imaging and inversion with prestack seismic data

The sparse expression of seismic data is important to signal processing, migration and parameter inversion. The conventional linear Radon transform is generalized to invert for both the local plane-wave and shot/receiver ray parameters simultaneously, which can compress the prestack seismic data in the local plane-wave domain. Consequently, it will be convenient to the subsequent ray-based imaging and inversion.Based on the combination of shot/receiver ray parameters, a new plane-wave prediction model is presented which can predict the local plane-wave within the spatial window of local shot/receiver center. The basic assumption is the local linear property of the wave-front. The seismic traces within the spatial window of local shot/receiver center are time-shifted and stacked according to the shot/receiver ray parameters and spatial distance, and the seismic phase with the given shot/receiver ray parameter can be separated naturally. Since the presented time-shift and stack procedure relies on the propagation direction of local plane-wave, we call it characteristic wave decomposition (CWD). Besides, the inversion-based CWD under the framework of compressed sensing (CS) theory can invert for both the local plane-wave and shot/receiver ray parameters simultaneously. Due to the pre-estimated ray-parameters for shot/receiver center, the loop for shot/receiver ray-parameters is avoided. Therefore the efficiency of conventional ray-based migration can be increased remarkably, and the ellipsoid migration response in constant media can be simplified to a point response. The new imaging scheme is called the characteristic wave imaging (CWI).First, the proposed CWD method is tested with the Marmousi synthetic data. Numerical examples show that the seismic phases can be separated completely. Meanwhile, even if the original seismic traces are contaminated by severe random noise, the numerical results demonstrate the robustness of CWD method.Then, the CWI method is tested with the Waxian synthetic data. Due to the high S/N ratio, the seismic data are compressed considerably. Compared with Kirchhoff PSDM method, the imaging efficiency is increased over one order of magnitude. Due to the pre-estimated ray-parameters, the angle domain CIGs can be produced directly.Finally, the CWD method is applied to the separation of first-arrival waveform, serving as the input data for wave-equation based transmission traveltime tomography.The proposed CWD method can compress the prestack seismic data in the local plane-wave domain. Under the framework of compressed sensing theory, both the local plane-wave and shot/receiver ray parameters can be inverted simultaneously. However, the local linear property of wave-front is the basic assumption of CWD. Therefore, it is not suitable for complex topography. Besides, in case of sparse shot/receiver sampling, it is difficult to invert for the shot/receiver ray-parameter simultaneously. The double beam forming procedure in CWD will be degenerated to the classical beam forming method.The presented CWI scheme can map the local plane-wave into the model space with the pre-estimated shot/receiver ray-parameters, so that the imaging efficiency is increased and the migration noise is attenuated. Meanwhile, the angle domain CIGs can be produced easily. Combined with the angle-domain tomography theory, it may be beneficial to the automation of velocity model building procedure.