物探化探计算技术2025，Vol.47Issue(2)：169-178,10.DOI:10.12474/wthtjs.20240108-0001
基于HS-DnCNN网络的地震数据随机噪声去除
Removal of random noise in seismic data based on HS-DnCNN network
姜懿珈 1周怀来 2陈姝颖 1吕芬1
作者信息
- 1. 成都理工大学地球物理学院,成都 610059||成都理工大学地球探测与信息技术教育部重点实验室,成都 610059
- 2. 成都理工大学地球物理学院,成都 610059||成都理工大学油气藏地质及开发工程国家重点实验室,成都 610059||成都理工大学地球探测与信息技术教育部重点实验室,成都 610059
- 折叠
摘要
Abstract
Seismic data,a crucial tool for exploration and monitoring,often faces interference from various noises,impacting the quality and accuracy of data interpretation.This study aims to improve the existing noise suppression network,Denoising Convolutional Neural Network(DnCNN),and enhance its ability to remove random noise from seismic data.In the denoising experiments,we extensively explore the design and performance influences of DnCNN.We systematically evaluate their performance in seismic data denoising by contrasting common activation functions,including traditional ReLU and Hard-swish.Results indicate that the HS-DnCNN network,constructed with Hard-swish,outperforms the original DnCNN in suppressing random noise.Moreover,compared to other denoising methods,it more effectively mitigates random noise.Subsequently,we apply HS-DnCNN to practical seismic data denoising experiments,validating its excellent applicability in real-world scenarios.HS-DnCNN demonstrates superior denoising effects while preserving and emphasizing essential signals.This summarizes the essence of our study.关键词
地震数据去噪/随机噪声/Hard-swish激活函数/DnCNNKey words
seismic data denoising/random noise/Hard-swish activation function/DnCNN引用本文复制引用
姜懿珈,周怀来,陈姝颖,吕芬..基于HS-DnCNN网络的地震数据随机噪声去除[J].物探化探计算技术,2025,47(2):169-178,10.
