Eliminating harmonic noise in vibroseis data through sparsity-promoted waveform modeling

Abstract

Vibroseis acquisition, which uses slip sweep instead of traditional flip-flop acquisition, could significantly reduce cycle time and increase productivity. However, the vibroseis system suffers from harmonically distorted sweeps being used as correlation operators, thus causing sticky harmonic distortions in correlated data that cannot be eliminated by forerunning manipulations and hindering interpretation. We propose a novel method to separate the harmonic interferences from correlated vibroseis data by exploring the waveform diversity between useful reflections and harmonic interferences. Following the diverse time-frequency distribution patterns of useful signal components and harmonic interferences, two different redundant waveform dictionaries are constructed to sparsely model useful reflections and harmonic interferences. Then, an iterative thresholding algorithm is used to gradually separate harmonic interferences from useful reflections, with each successive iteration potentially extracting the most reliable waveform elements built up into the corresponding signal components. The processing results of synthetic and field data examples highlight the effectiveness of our method in eliminating harmonic noise without noticeable loss of useful reflections. Compared to the classic frequency-dependent attenuation method, our approach has a higher fidelity.