Deep Learning Based Vz Noise Attenuation For OBS Data
Authors: Jing Sun, Arash JafarGandomi, Julian Holden
In ocean bottom seismic (OBS), combining records from the hydrophone (P-component) and vertical geophone (Z-component) allows for the elimination of receiver-side ghost and water-layer multiples. Apart from environmental noise that affects both of them, the Z-component is often contaminated by an additional type of shear-like noise known as Vz noise. It can cause significant issues with subsequent processing. Machine learning methods such as clustering seismic attributes extracted from the P- component in the Radon domain have been used for this task. We propose a deep learning (DL) approach.