AUTOMATIC ESTIMATION OF STACKING VELOCITY BASED ON SPARSE INVERSION

Abstract

Stacking velocity analysis is a routine procedure in seismic data processing, and is also a classical method for initial velocity model building. Usually, stacking velocity analysis is divided into two steps: calculating stacking velocity spectra and picking spectra maximums. Until now, many researchers are trying to improve the stacking velocity spectra by computing a better semblance, considering the AVO effect or improving the anti-noise ability of algorithm. However, it is seldom discussed on how to calculate the stacking velocity automatically. In this paper, we try to solve this problem by combining the velocity spectra calculation and picking procedure into a model parameter estimation under the framework of sparse inversion. Therefore, it is possible to invert the stacking velocity automatically and shorten the turn-around time of initial velocity model building and reduce human costs considerably. To solve this problem, first we give the definition of forward problem, which is the prediction model for CMP gather using stacking velocity and t₀ time as model parameters. Then, the inverse problem is defined as finding the sparse model parameters with the given CMP gather. Using the sparsity of model parameters as model constraint, we reformulate the conventional stacking velocity analysis problem as a new sparse inverse problem, and present an adaptive matching pursuit (MP) algorithm to solve it. The proposed method is quite promising for automatic initial model building, and can provide a good initial model for subsequent high-resolution velocity inversion methods. Numerical and field data tests demonstrate the effectiveness of the proposed method.