基于地震波场特征表达的数据域波动方程反射层析反演方法研究

目前，（各向异性）逆时偏移（RTM）技术和全波形反演（FWI）的技术组合成为了勘探方法的发展导向，在实际应用中取得了巨大的成功。然而，这些高精度偏移成像或反演成像方法严重依赖背景模型的精度，而背景模型的建立主要依赖地震波的走时信息。传统的射线层析反演方法（如成像道集层析）受制于射线理论的局限性，而基于波动方程的走时反演方法难以处理反射地震数据（难点在于反射时差的准确测量）。本项目从地震数据的特征表达出发，首先基于稀疏表达及字典学习等理论，对复杂波场中带限子波的走时信息进行特征表达与提取。接着，基于现代图像分析理论，发展一种稳健且不受振幅因素影响的带限地震信号反射时差自动测量方法。最后，构造一种波动方程反射走时反演方法，在数据域利用纯走时信息及全波算子反演背景速度模型，进一步提高背景速度模型的估计精度，为后续的高精度偏移成像或反演成像提供可靠的背景速度。

The combination of anisotropic reverse time migration (RTM) and full waveform inversion (FWI) has become a hot spot of exploration seismology, which has been succeed in many commercial applications. However, those high-resolution migration/inversion methods rely on the accuracy of background model, which should be inferred from the traveltime information of seismic data. The traveltime inversion method, such as the conventional ray-based tomographic method, suffers from the limitation of ray theory, while the wave-equation-based traveltime inversion method is difficult to apply to reflection seismic data due to the problem of reflection time-shift measurement. This research will start from the characteristic expression of seismic signal. First, based on the sparse representation and dictionary learning, the traveltime information of seismic data is extracted. Then, based on modern imaging processing theory, a stable method to calculate the reflection traveltime shift is presented. Finally, a wave-equation-based traveltime inversion method is proposed, which can further improve the accuracy of background velocity model, and provide a reliable initial model for subsequent high-resolution migration and inversion methods.