《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (3): 832-839.DOI: 10.11772/j.issn.1001-9081.2024101538

• 大模型前沿研究与典型应用 • 上一篇    下一篇

基于视觉大模型隐私保护的监控图像定位

李强1, 白少雄1, 熊源2, 袁薇3()   

  1. 1.国能运输技术研究院有限责任公司,北京 100080
    2.中山大学深圳校区 网络空间安全学院,广东 深圳 518107
    3.中国软件评测中心(工业和信息化部软件与集成电路促进中心),北京 102206
  • 收稿日期:2024-10-31 修回日期:2024-12-02 接受日期:2024-12-04 发布日期:2025-01-06 出版日期:2025-03-10
  • 通讯作者: 袁薇
  • 作者简介:李强(1996—),男,山西神池人,工程师,硕士,主要研究方向:智慧化系统建设、智能装备革新、安全运营
    白少雄(1997—),男,河北定州人,硕士,主要研究方向:计算机视觉
    熊源(1988—),男,湖北蕲春人,博士,主要研究方向:计算机视觉、计算机图形学、深度学习

Privacy preserving localization of surveillance images based on large vision models

Qiang LI1, Shaoxiong BAI1, Yuan XIONG2, Wei YUAN3()   

  1. 1.China Energy Institute of Transportation Technology Research Company Limited,Beijing 100080,China
    2.School of Cyber Science and Technology,SUN Yat-sen University,Shenzhen,Shenzhen Guangdong 518107,China
    3.China Software Testing Center (Ministry of Industry and Information Technology Software and Integrated Circuit Promotion Center),Beijing 102206,China
  • Received:2024-10-31 Revised:2024-12-02 Accepted:2024-12-04 Online:2025-01-06 Published:2025-03-10
  • Contact: Wei YUAN
  • About author:LI Qiang, born in 1996, M. S., engineer. His research interests include intelligent system construction, intelligent device update, operation safe.
    BAI Shaoxiong, born in 1997, M. S. His research interests include computer vision.
    XIONG Yuan, born in 1988, Ph. D. His research interests include computer vision, computer graphics, deep learning.

摘要:

监控图像的视觉定位是工业智能领域的关键技术。针对现有视觉定位算法缺少对图像中隐私信息的保护,在数据传输过程中容易导致敏感内容泄露的问题,提出一种基于视觉大模型(LVM)的监控图像定位方法。首先,设计基于LVM隐私保护的视觉定位架构,以利用少量文本提示和参考图像对输入图像进行风格迁移;其次,提出面向风格迁移图像的特征匹配算法用于相机位姿的估计。在公开数据集上的实验结果表明,所提方法的定位结果误差较小,在保证定位精度的前提下大幅减少了隐私泄露。

关键词: 扩散模型, 监控定位, 视觉大模型, 视觉定位, 隐私保护

Abstract:

Visual localization of surveillance images is an important technology in industrial intelligence. The existing visual localization algorithms lack the protection of the privacy information in the image and may lead to the leakage of sensitive content during data transmission. To address the problem, a localization method of surveillance images based on Large Vision Models (LVMs) was proposed. Firstly, the architecture of LVM privacy preserving-based visual localization was designed to transfer the style of input images by using a few prompts and reference images. Then, a feature matching algorithm for the image with style transfer was designed to estimate the camera pose. Experimental results on public datasets show that the localization error of the proposed algorithm is relatively small, demonstrating that the algorithm reduces the privacy leakage significantly while ensuring the localization accuracy.

Key words: diffusion model, surveillance localization, Large Vision Model (LVM), visual localization, privacy preserving

中图分类号: