Deep learning for ground-roll noise attenuation

Abstract

Deep learning, an advance machine learning method, has gained a lot of interest in various fields in recent years. In this paper, deep learning is investigated to suppress pre-stack seismic noise by learning the features of noise in seismic data automatically. First of all, a convolutional neural network (CNN) architecture for ground-roll noise attenuation is designed, then the network is trained by using the scattered ground-roll noise obtained from two shots in advance as labels, and finally, the denoising network is applied to all shots of the field land shot gather. The feasibility of this approach is confirmed by the denoising results, and this approach is shown to be promising in suppressing the scattered ground-roll noise automatically with high computational efficiency. Presentation Date: Tuesday, October 16, 2018 Start Time: 1:50:00 PM Location: 204B (Anaheim Convention Center) Presentation Type: Oral