A new method for seismic noise removal based on deep learning consistently produces superior results to traditional noise attenuation processes. The fully automated noise attenuation method has been developed that eliminates the need for time-consuming and subjective user parameter testing.
Both instrumental and environmental noise needs to be removed from motion sensor records in multisensor streamer acquisition. The new method is outlined in a PGS First Break paper in February and consists of two convolutional neural network models. The first model attenuates vertical narrow band high amplitude noise mainly generated by the instruments attached to the streamers such as eBirds.










