基于稀疏重建分析的道集优化方法

地震叠前道集处理的优劣一方面决定了叠后处理结果的好坏, 另一方面影响地震AVO解释结果的正确与否, 因此叠前道集的优化处理具有重要的意义。受子波拉伸、拉伸调谐效应、薄层调谐效应等因素的影响, 叠前道集记录包含了除地层反射信息以外的无用的干涉信息, 单一子波字典不能对其进行很好的稀疏表示, 而过完备子波字典能够对地震信号进行多特征稀疏表示。针对共反射点(CRP)叠前道集, 研究了基于稀疏重建策略的道集优化方法。首先, 建立了地震记录子波投影下的稀疏反射系数估计的约束优化目标函数和基于最小二乘的无约束优化目标函数, 为了保障解的稳定性, 引入了正则化策略和震荡平滑策略。其次, 在雷克子波字典库基础上, 考虑主频、吸收衰减、时移及相位变化等因素, 建立多特征表示的复杂过完备字典库, 地震记录通过过完备子波字典投影后, 反射系数、子波的拉伸效应、调谐效应等能够较好地进行稀疏表示, 进而提取更准确的反射系数。综合以上两方面的研究, 形成了过完备字典库表示下的正则化正交匹配追踪算法。理论模型和实际资料的处理结果表明, 处理后的地震叠前道集在噪声压制、拉伸校正及提高分辨率等方面都有一定程度的改善与优化。

As the quality of pre-stack processing determines the quality of post-stack results and affects the accuracy of seismic AVO interpretation, improvement and optimization of the pre-stack trace set is essential.Owing to the influence of wavelet stretching, stretching tuning effect, thin layer tuning effect, and other factors, it contains redundant interference information except for formation reflection, which further complicates pre-stack track set recording.A single wavelet dictionary cannot represent the seismic signal effectively, whereas an overcomplete wavelet dictionary can perform multifeatured sparse representations of seismic signals.For a common reflection point pre-stack trace set, a path set optimization method based on a sparse reconstruction strategy was studied.Initially, the constrained optimization objective function and unconstrained optimization objective function based on least square were established for sparse reflection coefficient estimation under the wavelet projection of seismic records.Subsequently, regularization and oscillation smoothing strategies were introduced to ensure the stability of the solution.Based on the built Ricker wavelet dictionary and considering the main frequency, absorption attenuation, and factors such as time shift and phase changes, a more complex over complete dictionary library feature was constructed.After complete wavelet dictionary projection, the reflection coefficient, tensile effect of wavelet, composition, such as tuning effect of the seismic data, could be effectively represented, and a more accurate reflection coefficient could be extracted.From the above results, a regularized orthogonal matching tracking algorithm based on an over-complete dictionary representation was developed.The results of the theoretical model and practical data show that noise suppression, tensile correction, and resolution enhancement were improved and optimized to a certain extent.