计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (6A): 250400077-7.doi: 10.11896/jsjkx.250400077

• 图像处理&多媒体技术 • 上一篇    下一篇

基于改进MobileNetV4的电力设备红外与可见光图像单应性估计

王胜1,2, 张凌浩1,2, 张菊玲1,2, 庞博1,2, 郗宁3, 佘文魁4   

  1. 1 国网四川省电力公司电力科学研究院 成都 610043
    2 新型电力系统安全与运行四川省重点实验室 成都 610043
    3 国网天府新区供电公司 成都 610200
    4 四川中电启明星信息技术有限公司 成都 610200
  • 出版日期:2026-06-16 发布日期:2026-06-12
  • 通讯作者: 佘文魁(shewenkui@126.com)
  • 作者简介:(w1121s@163.com)
  • 基金资助:
    国网四川省电力公司科技项目(52199723002P)

Infrared and Visible Image Homography Estimation for Power Equipment Based on Improved MobileNetV4

WANG Sheng1,2, ZHANG Linghao1,2, ZHANG Juling1,2, PANG Bo1,2, XI Ning3, SHE Wenkui4   

  1. 1 State Grid Sichuan Electric Power Research Institute,Chengdu 610043,China
    2 Power System Security and Operation Key Laboratory of Sichuan Province,Chengdu 610043,China
    3 State Grid Tianfu New Area Power Supply Company,Chengdu 610200,China
    4 Aostar Information Technology Co.,Ltd.,Chengdu 610200,China
  • Published:2026-06-16 Online:2026-06-12
  • About author:WANG Sheng,born in 1987,master,senior engineer.His main research interests include power grid network security attack and defense technology,data security,Internet of things security,and industrial control security.
    SHE Wenkui,born in 1982,master,se-nior engineer.His main research in-terests include cloud computing,power Internet of Things.
  • Supported by:
    Scientific Research Foundation of State Grid Sichuan Electric Power Company(52199723002P).

PDF (PC)

32

摘要： 红外图像与可见光图像的单应性估计是提升电力设备定位精度和缺陷检测准确度的关键技术之一。针对现有方法在电力设备红外与可见光图像单应性估计中精度不足和模型规模较大的问题,提出了一种轻量化的基于改进MobileNetV4的单应性估计方法。首先,首次将MobileNet应用于单应性估计任务,设计了一种轻量级的估计模型。其次,通过在MobileNetV4的各阶段引入CBAM模块,突出了特征图中的关键特征,从而提出了一种改进的MobileNetV4模型,即CBMobileNet。最后,使用L1范数剪枝算法,在确保性能损失较小的同时,大幅降低了模型的参数量和计算复杂度。实验结果表明,在合成基准数据集上,相较于次优算法,所提方法的平均角点误差从5.06显著下降至4.95。此外,相较于原始模型,剪枝后的模型在参数量上从10.04 MB显著减少至6.91 MB,FLOPs从1 029.48 MB显著降低至755.11 MB,而平均角点误差仅从4.93略微上升至4.95。

关键词: 单应性估计, MobileNetV4, 模型剪枝, 红外与可见光图像, 电力设备

Abstract: The homography estimation of infrared and visible images is one of the key techniques to improve the positioning accuracy and defect detection accuracy of power equipment.To address the problems of insufficient accuracy and large model size of existing methods in homography estimation of infrared and visible images of power equipment,a lightweight homography estimation method based on improved MobileNetV4 is proposed.Firstly,MobileNet is applied to the homography estimation task for the first time,and a lightweight estimation model is designed.Secondly,an improved MobileNetV4 model,CBMobileNet,is proposed by highlighting the key features in the feature map through the introduction of the CBAM module in each stage of MobileNetV4.Finally,the number of parameters and computational complexity of the model is significantly reduced using the L1 Norm pruning algorithm while ensuring less performance loss.The experimental results show that the average corner error of the proposed method substantially decreases from 5.06 to 4.95 compared to the suboptimal algorithm on the synthetic benchmark dataset.In addition,compared to the original model,the pruned model significantly reduces the parameters from 10.04 MB to 6.91 MB and the FLOPs from 1 029.48 MB to 755.11 MB,while the average corner error only slightly increases from 4.93 to 4.95.

Key words: Homography estimation, MobileNetV4, Model pruning, Infrared and visible image, Power equipment

中图分类号: 

  • TP391.41
[1] FU T,DENG C Z,HAN X Y,et al.Infrared and Visible Image Registration for Power Equipments Based on Deep Learnin[J].Infrared Technolog,2022,44(9):936-943.
[2] WEN S Y,ZHANG S,ZHANG C Y,et al.Multi-scale object detection algorithm based on improved YOLOv8[J].Modern Electronics Techniqu,2024,47(15):133-138.
[3] RAHMAN E U,ZHANG Y,AHMAD S,et al.Autonomous vision-based primary distribution systems porcelain insulators inspection using UAVs[J].Sensors,2021,21(3):974.
[4] LIN Y,LIU M,BAI D M,et al.Infrared-visible Image Smart Registration for Electrical Equipment Based on Deep Learning [J].Shandong Electric Power,2022,49(8):22-27.
[5] WANG X,LUO Y,FU Q,et al.Coarse-to-Fine HomographyEstimation for Infrared and Visible Images[J].Electronics,2023,12(21):4441.
[6] LOWE D G.Distinctive image features from scale-invariant keypoints[J].International journal of computer vision,2004,60:91-110.
[7] BAY H,TUYTELAARS T,VAN GOOL L.Surf:Speeded up robust features[C]//Computer Vision－ECCV 2006:9th European Conference on Computer Vision.Springer,2006:404-417.
[8] RUBLEE E,RABAUD V,KONOLIGE K,et al.ORB:An efficient alternative to SIFT or SURF[C]//2011 International conference on computer vision.IEEE,2011:2564-2571.
[9] WANG X Y,LUO Y H,WU Y Z,et al.Homography Estimation of Infrared and Visible Images Based on Unet[J].Aeronautical Computing Technique,2023,53(4):51-55.
[10] WANG X Y.Research on Homography Estimation for Infra-redand Visible Images based on Deep Learning [D].Guanghan:Civil Aviation Flight University of China,2024.
[11] LUO Y,WANG X,WU Y,et al.Detail-Aware Deep Homography Estimation for Infrared and Visible Image[J].Electronics,2022,11(24):4185.
[12] LUO Y,WANG X,WU Y,et al.Infrared and visible image homography estimation using multiscale generative adversarial network[J].Electronics,2023,12(4):788.
[13] WANG X,LUO Y,FU Q,et al.Infrared and Visible Image Homography Estimation Based on Feature Correlation Transformers for Enhanced 6G Space-Air-Ground Integrated Network Perception[J].Remote Sensing,2023,15(14):3535.
[14] QIN D,LEICHNER C,DELAKIS M,et al.MobileNetV4-Universal Models for the Mobile Ecosystem[J].arXiv:240410518,2024.
[15] WOO S,PARK J,LEE J Y,et al.Cbam:Convolutional block attention module[C]//Proceedings of the European conference on computer vision(ECCV).2018:3-19.
[16] LI H,KADAV A,DURDANOVIC I,et al.Pruning filters for efficient convnets[J].arXiv:160808710,2016.
[17] GKRISPANIS K,GKALELIS N,MEZARIS V.Filter-Pruningof Lightweight Face Detectors Using a Geometric Median Criterion[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision.2024:280-289.
[18] JIANG Q,LIU Y,YAN Y,et al.A contour angle orientation for power equipment infrared and visible image registration[J].IEEE Transactions on Power Delivery,2020,36(4):2559-2569.
[19] ZHANG J,WANG C,LIU S,et al.Content-aware unsupervised deep homography estimation[C]//Computer Vision－ECCV 2020:16th European Conference.Springer,2020:653-669.
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