Seismic signal fast decomposition by multichannel matching pursuit with Genetic Algorithm

Abstract

The atoms obtained by single-channel matching pursuit (SCMP) are unstable, and the solution is not unique, moreover, the reconstructed profiles lack lateral continuation. Multichannel matching pursuit (MCMP) can overcome the above shortcomings because it exploits lateral coherence as a constraint and decomposes all seismic signals at the same time and gets the best time-frequency atoms that be shared by all the seismic traces. But the existed MCMP algorithm is liable to being influenced by noise and has quite low computing efficiency. This paper proposes an MCMP algorithm that decomposes all seismic signals at the same time through changing the criterion function with Genetic Algorithm. The experiments for both synthetic and actual seismic data show that our algorithm can obtain stable atoms, reduce the influence of data noise to some extent, and improve computing efficiency by more than 100 times.