Full text loading...
-
Motion Sensor Noise Attenuation Using Deep Learning
- Source: First Break, Volume 41, Issue 2, Feb 2023, p. 45 - 51
-
- 01 Feb 2023
Abstract
A method is proposed to attenuate both instrumental and environmental noise from motion sensor records in multisensor streamer acquisition. The main elements are two convolutional neural network models. The first model attenuates vertical narrow band high amplitude noise mainly generated by the instruments attached to the streamers. The second model attenuates widespread background noise mainly associated with environmental conditions. To reduce the risk of possible signal loss an addback flow in the curvelet domain is used. The motivation for the work presented here was to develop a fully automated noise attenuation method that eliminates the need for time-consuming and subjective user parameter testing. The method has been validated using seismic data from different parts of the world and shown to consistently produce superior results to other state-of-the-art noise attenuation processes.