Reflection Waveform Inversion Regularized with Structural Constraints and Scattered Data Interpolation

Chengliang Wu,Huazhong Wang,Bo Feng,Ting Lei,Rongwei Xu,Shen Sheng

DOI: https://doi.org/10.1190/image2023-3911408.1

2023-01-01

Abstract:Reflection waveform inversion (RWI) is a well-known technique used to recover deep velocity models by utilizing reflection data. However, RWI is inherently an ill-posed inversion problem and incorporating prior information can help reduce the null space and improve the conditioning of the inversion. One effective approach is to introduce model regularization such as mathematical constraints (Total-Variation regularization) or structure-oriented constraints. Nonetheless, the effectiveness of the structure-oriented method heavily relies on the accuracy of the reflection angle and is limited to the vicinity of the reflection interface, resulting in reduced structural filtering capabilities when moving away from the reflection layer. In this abstract, we propose a novel method that combines structural constraints and scattered data interpolation (SDI) to enhance RWI by incorporating geological structures. Firstly, we employ the Markov decision process (MDP) to automatically extract fundamental reflection structures. Subsequently, the SDI method is employed to capture the structural characteristics between reflection layers. Moreover, we design a smoothing operator that conforms to the geological structure characteristics at both the reflection interface and non-structural locations. Lastly, we use synthetic data to verify the effectiveness of the proposed method.