计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (5): 193-206.doi: 10.11896/jsjkx.250400117

• 计算机图形学 & 多媒体 • 上一篇    下一篇

基于预训练网络的重力波光谱图像异常检测及定位技术

黄思扬1, 姚烨2, 朱怡安2, 海铎1, 熊志海1   

  1. 1 西北工业大学软件学院 西安 710072
    2 西北工业大学计算机学院 西安 710072
  • 收稿日期:2025-04-23 修回日期:2025-07-22 发布日期:2026-05-08
  • 通讯作者: 姚烨(yaoye@nwpu.edu.cn)
  • 作者简介:(hsy20000216@126.com)
  • 基金资助:
    国家重点研发计划(2021YFC2802503)

Anomaly Detection and Localization Technology for Gravity Wave Spectral Images Based onPre-trained Networks

HUANG Siyang1, YAO Ye2, ZHU Yian2, HAI Duo1, XIONG Zhihai1   

  1. 1 School of Software, Northwestern Polytechnical University, Xi’an 710072 , China
    2 School of Computing, Northwestern Polytechnical University, Xi’an 710072, China
  • Received:2025-04-23 Revised:2025-07-22 Online:2026-05-08
  • About author:HUANG Siyang,born in 2000,postgra-duate.His main research interests include digital twin and image anomaly detection.
    YAO Ye,born in 1972,associate professor.His main research interests include network information security,digital twin technology,system health management,operation and maintenance,etc.
  • Supported by:
    National Key Research and Development Program of China(2021YFC2802503).

PDF (PC)

237

摘要： 针对重力波光谱图像数据特征不明显、分布不均,容易导致异常检测误差高等问题,提出一种基于预训练网络的重力波光谱图像异常检测方法。首先,通过对重力波光谱图像特征进行分析,利用预处理方法增强图像的关键特征,来更准确地获得光谱图像中的有用特征信息;然后,采用ImageNet预训练网络的中间层进行特征提取,通过核心集子采样机制压缩特征内存库,以缩短推理分析时间;最后,采用最近邻机制计算光谱图像像素点的异常分值,并据此实现对重力波光谱图像整体异常程度的评估以及异常区域的标定。实验结果表明,所提方法能够细粒度地进行重力波光谱图像特征分析,有效利用图像特征进行异常检测,并对重力波光谱图像异常区域进行精确标定,异常判别和定位AUROC指标分别达到98.73%和95.19%。

关键词: 异常检测, 核心集子采样, 无监督学习, 预训练模型, 重力波光谱图像

Abstract: To addresses the issues of indistinct features and uneven distribution in gravitational wave spectral image data,which often lead to high error rates in anomaly detection.This paper proposes an anomaly detection method for gravitational wave spectral images based on a pre-trained network.This method analyzes image features at both the image level and pixel level,employing preprocessing techniques to enhance the key features of the images,thereby more accurately capturing useful feature information within the images.The intermediate layers of an ImageNet pre-trained network are utilized for feature extraction,and a core-set subsampling mechanism is applied to compress the feature memory bank,reducing inference analysis time.Finally,the nearest neighbor algorithm is used to calculate the anomaly scores of image pixels,enabling the assessment of the overall anomaly degree of the image and the identification of anomalous regions.Experimental results demonstrate that this method can effectively analyze features of gravitational wave spectral images at both image and pixel levels,utilize image features for anomaly detection,and accurately identify anomalous regions in gravitational wave spectral images.The AUROC metrics for anomaly discrimination and localization reach 98.73% and 95.19%.

Key words: Anomaly detection, Core subset sampling, Unsupervised learning, Pretrained model, Gravity wave spectral image

中图分类号: 

  • TP391
[1]XIE Z,ZHANG P,WU Z,et al.Legacy and emerging organiccontaminants in the polar regions[J].Science of the Total Environment,2022,835:155376.
[2]GÓMEZ Á L P,MAIMÓ L F,CELDRÁN A H,et al.SUSAN:A Deep Learning based anomaly detection framework for sustainable industry[J].Sustainable Computing:Informatics and Systems,2023,37:100842.
[3]BOGDOLL D,NITSCHE M,ZÖLLNER J M.Anomaly detection in autonomous driving:A survey[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:4488-4499.
[4]YANG J,XU R,QI Z,et al.Visual anomaly detection for images:A systematic survey[J].Procedia Computer Science,2022,199:471-478.
[5]IQBAL H,KHALID U,CHEN C,et al.Unsupervised anomaly detection in medical images using masked diffusion model[C]//International Workshop on Machine Learning in Medical Imaging.Cham:Springer,2023:372-381.
[6]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Imagenetclassification with deep convolutional neural networks[C]//Advances in Neural Information Frocessing Systems.2012.
[7]DENG J,DONG W,SOCHER R,et al.Imagenet:A large-scale hierarchical image database[C]//2009 IEEE Conference on Computer Vision and Pattern Recognition,2009:248-255.
[8]COHEN N,HOSHEN Y.Sub-image anomaly detection withdeep pyramid correspondences[J].arXiv:2005.02357,2020.
[9]BERGMANN P,FAUSER M,SATTLEGGER D,et al.MVTec AD--A comprehensive real-world dataset for unsupervised anomaly detection[C]//Proceedings of the IEEE/CVF Confe-rence on Computer Vision and Pattern Recognition.2019:9592-9600.
[10]XIANG P,ALI S,JUNG S K,et al.Hyperspectral anomaly detection with guided autoencoder[J].IEEE Transactions on Geoscience and Remote Sensing,2022,60:1-18.
[11]CHEN Z,YEO C K,LEE B S,et al.Autoencoder-based network anomaly detection[C]//2018 Wireless Telecommunications Symposium(WTS).IEEE,2018:1-5.
[12]BAUR C,WIESTLER B,ALBARQOUNI S,et al.Deep autoencoding models for unsupervised anomaly segmentation in brain MR images[C]//Brainlesion:Glioma,Multiple Sclerosis,Stroke and Traumatic Brain Injuries:4th International Workshop.Springer,2019:161-169.
[13]AKCAY S,ATAPOUR-ABARGHOUEI A,BRECKON T P.Ganomaly:Semi-supervised anomaly detection via adversarial training[C]//Computer Vision-ACCV 2018:14th Asian Conference on Computer Vision.Springer,2019:622-637.
[14]AKÇAY S,ATAPOUR-ABARGHOUEI A,BRECKON T P.Skip-ganomaly:Skip connected and adversarially trained encoder-decoder anomaly detection[C]//2019 International Joint Conference on Neural Networks(IJCNN).IEEE,2019:1-8.
[15]ZENATI H,FOO C S,LECOUAT B,et al.Efficient gan-based anomaly detection[J].arXiv:1802.06222,2018.
[16]ZENATI H,ROMAIN M,FOO C S,et al.Adversarially learned anomaly detection[C]//2018 IEEE International Conference on Data Mining(ICDM).IEEE,2018:727-736.
[17]PIDHORSKYI S,ALMOHSEN R,DORETTO G.Generativeprobabilistic novelty detection with adversarial autoencoders[C]//Advances in Neural Information Processing Systems.2018.
[18]CHEN C,CHEN P,SONG H,et al.Anomaly detection by one class latent regularized networks[J].arXiv:2002.01607,2020.
[19]YI J,YOON S.Patch svdd:Patch-level svdd for anomaly detection and segmentation[C]//Proceedings of the Asian Conference on Computer Vision.2020.
[20]ZHENG Y,WANG X,DENG R,et al.Focus your distribution:Coarse-to-fine non-contrastive learning for anomaly detection and localization[C]//2022 IEEE International Conference on Multimedia and Expo(ICME).IEEE,2022:1-6.
[21]DEFARD T,SETKOV A,LOESCH A,et al.Padim:a patch distribution modeling framework for anomaly detection and localization[C]//International Conference on Pattern Recognition.Cham:Springer,2021:475-489.
[22]LIRON B,NIV C,YEDID H.Deep nearest neighbor anomaly detection[J].arXiv:2002.10445,2020.
[23]CIREGAN D,MEIER U,SCHMIDHUBER J.Multi-columndeep neural networks for image classification[C]//2012 IEEE Conference on Computer Vision and Pattern Recognition.IEEE,2012:3642-3649.
[24]CIREŞAN D C,MEIER U,MASCI J,et al.High-performance neural networks for visual object classification[J].arXiv:1102.0183,2011.
[25]HINTON G E,SRIVASTAVA N,KRIZHEVSKY A,et al.Improving neural networks by preventing co-adaptation of feature detectors.arXiv 2012[J].arXiv:1207.0580,2012.
[26]JARRETT K,KAVUKCUOGLU K,RANZATO M A,et al.What is the best multi-stage architecture for object recognition?[C]//2009 IEEE 12th International Conference on Computer Vision.IEEE,2009:2146-2153.
[27]LI L C,GAO H Y,BU L B,et al.Research on the inversion of rotational temperature of gas glow layer based on O2(0-1) spectral band[J].Spectroscopy and Spectral Analysis,2020,40(10):3002-3009.
[28]LU C T,KOU Y,ZHAO J,et al.Detecting and tracking regional outliers in meteorological data[J].Information Sciences,2007,177(7):1609-1632.
[29]RANCIC D D,EFERICA P M,DJORDJEVIC-KAJAN S J,et al.Digital signal and image processing for meteorological radars[C]//2000 10th Mediterranean Electrotechnical Confe-rence.Information Technology and Electrotechnology for the Mediterranean Countries.IEEE,2000:623-626.
[30]ZAGORUYKO S.Wide residual networks[J].arXiv:1605.07146,2016.
[31]HAR-PELED S,MAZUMDAR S.On coresets for k-means and k-median clustering[C]//Proceedings of the Thirty-sixth Annual ACM Symposium on Theory of Computing.2004:291-300.
[32]OZAN S,SILVIO S.Active learning for convolutional neuralnetworks:A core-set approach[C]//International Conference on Learning Representations.2018.
[33]SAMARTH S,HAN Z,ANIRUDH G,et al.Small-GAN:Speeding up GAN training using core-sets[C]//Proceedings of the 37th International Conference on Machine Learning.2020:9005-9015.
[34]SANJOY D,ANUPAM G.An elementary proof of a theorem of johnson and lindenstrauss[J].Random Structures & Algorithms,2003,22(1):60-65.
[35]GUDOVSKIY D,ISHIZAKA S,KOZUKA K.Cflow-ad:Real-time unsupervised anomaly detection with localization via conditional normalizing flows[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Cision.2022:98-107.
[36]ZAVRTANIK V,KRISTAN M,SKOČAJ D.Draem-a discriminatively trained reconstruction embedding for surface anomaly detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2021:8330-8339.
[37]LEE S,LEE S,SONG B C.Cfa:Coupled-hypersphere-based feature adaptation for target-oriented anomaly localization[J].IEEE Access,2022,10:78446-78454.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发