Morten W. Pedersen | Convolutional Neural Networks

13 October | 14:15 - 14:40 CST (GMT-5). Application of a convolutional neural network to classification of swell noise attenuation.

 

Session: SPET P3 Signal Processing and Imaging

Authors: Bagher Farmani and Morten W. Pedersen, PGS.

Morten Pedersen builds upon previously published work on the use of Convolutional Neural Networks (CNNs) for seismic noise removal. A mixed class classification approach with a U-net image segmentation model improves the classification of swell noise, and a strategy based on differential noise levels measured across the class levels robustly detects signal leakage in the attenuated noise. It is demonstrated how the improved classification of noise and leakage also improves the proposed automated swell noise removal.