Two objective matching pursuit for seismic data decomposition

Matching pursuit is an effective method for seismic data decomposition. The traditional matching pursuit has one single objective that successively maximizes the inner products between signal residuals and atoms at different approximating levels. However, such inner product strategy does not guarantee the consistency of the peak of an atom and that of its corresponding signal residual. To address this limitation, we propose a two objective matching pursuit strategy that not only optimizes the traditional matching pursuit but also minimizes peak distance. Compared with the traditional matching pursuit with one single objective, the two objective matching pursuit has higher search ability. In this scenario, the genetic algorithm (GA) used for solving the traditional matching pursuit is not suitable for the two objective matching pursuit. Therefore, we propose to solve the two objective optimization problem by exploiting a nondominated sorting genetic algorithm (NSGA-II) algorithm, which not only increases the diversity of searching optimization and but also reduces the reconstruction error over the GA. Experiments for seismic data validate the advantages of the proposed two objective matching pursuit.

2019 IEEE Symposium Series on Computational Intelligence (SSCI)