《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (10): 3336-3341.DOI: 10.11772/j.issn.1001-9081.2024101545

• 前沿与综合应用 • 上一篇    

基于多级小波残差网络的重力数据去噪方法

薛雅丽1(), 徐忠敏1,2, 刘世豪3   

  1. 1.南京航空航天大学 自动化学院,南京 211106
    2.常州慧勤新能源科技有限公司,江苏 常州 213164
    3.北京电子工程总体研究所,北京 100854
  • 收稿日期:2024-10-30 修回日期:2024-12-25 接受日期:2024-12-30 发布日期:2025-01-06 出版日期:2025-10-10
  • 通讯作者: 薛雅丽
  • 作者简介:薛雅丽(1974—),女,黑龙江集贤人,副教授,博士,主要研究方向:目标检测与识别、图像对抗与防御 Email:xueyali@nuaa.edu.cn
    徐忠敏(1988—),男,江苏沛县人,助理工程师,主要研究方向:发动机控制
    刘世豪(1997—),男,湖南安化人,助理工程师,主要研究方向:指挥控制与通信。
  • 基金资助:
    上海航天科技创新基金资助项目(SAST2022-013)

Gravity data denoising method based on multilevel wavelet residual network

Yali XUE1(), Zhongmin XU1,2, Shihao LIU3   

  1. 1.College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing Jiangsu 211106,China
    2.Huiqin (Changzhou) New Energy Technology Company Limited,Changzhou Jiangsu 213164,China
    3.Beijing Institute of Electronic System Engineering,Beijing 100854,China
  • Received:2024-10-30 Revised:2024-12-25 Accepted:2024-12-30 Online:2025-01-06 Published:2025-10-10
  • Contact: Yali XUE
  • About author:XUE Yali, born in 1974, Ph. D., associate professor. Her research interests include target detection and recognition, image countermeasures and defense.
    XU Zhongmin, born in 1988, assistant engineer. His research interests include engine control.
    LIU Shihao, born in 1997, assistant engineer. His research interests include command control and communications.
  • Supported by:
    Shanghai Aerospace Science and Technology Innovation Fund(SAST2022-013)

摘要:

为降低干扰噪声对重力实测数据的影响,进一步提高重力数据处理精度,提出一种基于多级小波残差网络(MWRNet)的重力数据去噪方法,该方法结合小波变换和神经网络实现对重力数据中噪声分量的去除。首先通过小波变换分解重力数据,再利用神经网络提取噪声,同时引入残差通道注意力(RCA)模块增强网络的噪声提取能力。利用模拟数据和实测数据测试所提方法,实验结果表明:所提方法相较于其他重力数据去噪算法具有更好的效果。在噪声水平为50的实验中,所提方法相较于传统去噪算法三维块匹配算法BM3D(Block-Matching and 3D filtering),在峰值信噪比(PSNR)、结构相似性指数(SSIM)上分别提升了21.8%、9.3%;相较于基于深度学习的去噪算法DnCNN(Denoising Convolutional Neural Network)、MWCNN(Multi-level Wavelet-CNN),PSNR、SSIM也分别有所提升。

关键词: 重力数据, 小波变换, 深度学习, 残差连接, 通道注意力

Abstract:

In order to reduce the influence of interference noise on gravity measured data and further improve the accuracy of gravity data processing, a gravity data denoising method based on Multilevel Wavelet Residual Network (MWRNet) was proposed, which combined wavelet transform and neural network to realize removal of noise components in gravity data. Firstly, the gravity data was decomposed by wavelet transform, and then a neural network was utilized for noise extraction, while the Residual Channel Attention (RCA) module was introduced to enhance noise extraction ability of the network. The proposed gravity data denoising method was tested using simulated data and measured data, and experimental results show that the proposed method has better results compared with other gravity data denoising algorithms. In specific, with noise level of 50, in Peak Signal-to-Noise Ratio (PSNR) and Structure SIMilarity (SSIM), the proposed method improves over 21.8% and 9.3%, respectively, compared to the traditional denoising algorithm BM3D (Block-Matching and 3D filtering). Compared to the deep learning-based denoising algorithms DnCNN (Denoising Convolutional Neural Network) and MWCNN (Multi-level Wavelet Convolutional Neural Network), PSNR and SSIM are improved respectively.

Key words: gravity data, wavelet transform, deep learning, residual connectivity, channel attention

中图分类号: