作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2023, Vol. 49 ›› Issue (4): 297-302,311. doi: 10.19678/j.issn.1000-3428.0066093

• 开发研究与工程应用 • 上一篇    下一篇

基于CSA-ResNet的人员入侵检测方法

张雷1,2, 鲍蓉1, 朱永红1, 史新国3   

  1. 1. 徐州工程学院 信息工程学院(大数据学院), 江苏 徐州 221000;
    2. 东南大学 移动通信国家重点实验室, 南京 210096;
    3. 山东能源淄博矿业集团有限公司信息中心, 山东 淄博 225100
  • 收稿日期:2022-10-24 修回日期:2022-11-25 发布日期:2022-12-13
  • 作者简介:张雷(1987-),男,讲师、博士,主研方向为移动无线感知;鲍蓉、朱永红(通信作者),教授、博士;史新国,高级工程师、硕士。
  • 基金资助:
    江苏省高等学校基础科学(自然科学)研究项目(21KJB510025);江苏省产学研合作项目(BY2021160);教育部产学合作协同育人项目(BY2021160202102356012);徐州市科技计划项目(KC19208);淄矿集团智慧矿山关键技术研发开放基金项目(2019LH05)。

Method for Detecting Personnel Intrusion Based on CSA-ResNet

ZHANG Lei1,2, BAO Rong1, ZHU Yonghong1, SHI Xinguo3   

  1. 1. School of Information Engineering (School of Big Data), Xuzhou University of Technology, Xuzhou 221000, Shandong, China;
    2. National Mobile Communications Research Key Laboratory, Southeast University, Nanjing 210096, China;
    3. Information Center, Shandong Energy Zibo Mining Group Co., Ltd., Zibo 225100, Shandong, China
  • Received:2022-10-24 Revised:2022-11-25 Published:2022-12-13

摘要: 视频监控作为最常用的监测方法,由于存在监控死角以及侵犯人员隐私等问题,存在许多应用瓶颈。针对视频监测无法用于敏感场景的问题,提出一种基于WiFi的人员入侵感知方法。该方法利用WiFi信号覆盖范围大、易获取的特点,实现无隐私侵犯与无死角监控。基于人员入侵对传输路径的影响,分析WiFi感知机理,建立基于WiFi状态信息的人员入侵检测感知模型,并设计子载波选择算法获取人员感知敏感子载波。通过离群点滤波、离散小波去噪等方法对采集的数据进行处理,根据人员入侵对信号的影响构造人员感知特征值。在此基础上,将特征信号和处理后的信道状态信息作为输入信息,放入基于通道和空间注意力残差网络的人员入侵检测模型中进行判识,并在多种场景下对该方法进行实验测试,分析影响检测精度的因素。实验结果表明,该方法在多种场景下平均检测准确率达到97.8%,能够满足多场景下人员入侵的检测要求。

关键词: 无线感知, 子载波选择算法, 人员入侵检测, 信道状态信息, 深度学习

Abstract: Video monitoring is the most widely used monitoring method, but it has many application bottlenecks owing to problems such as monitoring dead spots and invasion of personnel privacy.For problems such as video monitoring that cannot be used for sensitive scenes, a WiFi-based sensing method for detecting personnel intrusion is proposed.The method takes advantage of the wide coverage and easy access of WiFi signals to achieve privacy invasion-free and dead-end monitoring.First, based on the impact of person intrusion on transmission path, the WiFi sensing mechanism is analyzed, and a sensing model for person intrusion detection based on WiFi information state information is established. Next, a subcarrier selection algorithm is designed to obtain person-perception sensitive subcarriers.The collected data are processed by outlier point filtering and discrete wavelet denoising, and the person-perception feature values are constructed based on the impact of person intrusion on the signal.Finally, the feature signals and the processed channel state information are selected as input information and transmitted to the proposed channel and spatial attention residual network-based intrusion detection model for identification.The method is experimentally tested in multiple scenarios, and the factors affecting the detection accuracy are analyzed.The average accuracy is 97.8% in multiple scenarios, which satisfies the requirements of personnel intrusion detection in multiple scenarios.

Key words: wireless sensing, subcarrier selection algorithm, personnel intrusion detecting, Channel Status Information(CSI), deep learning

中图分类号: