Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
关于 JoVE
概览领导团队博客JoVE 帮助中心
作者
出版流程编辑委员会范围与政策同行评审常见问题投稿
图书馆员
用户评价订阅访问资源图书馆顾问委员会常见问题
研究
JoVE JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序
Same author

Investigation of proton spallation effect on Electron Emission Coefficient electrodes coated with metamaterial.

Applied radiation and isotopes : including data, instrumentation and methods for use in agriculture, industry and medicine·2026
Same author

Symmetry-guided explainable deep learning for colon cancer diagnosis: model benchmarking, cross-validation, statistical analysis, and explainability via ablation studies.

Frontiers in artificial intelligence·2026
Same author

Arsenic trioxide allosterically inhibits human telomerase, validated by in-silico and in-vitro cancer and non cancer cell lines.

Scientific reports·2026
Same author

SARS-CoV-2 Spike Protein S2 Subunit: Recombinant Protein Expression Analysis, Purification, and Its Regulatory Effect on IGF-1R Expression.

The protein journal·2026
Same author

SARS-CoV-2 Genome and S2 Spike Protein: IRF-Driven Interferon Regulation and Host Cell Responses.

Reviews in medical virology·2025
Same author

Production of secondary particles from cosmic ray interactions in the earth's atmosphere: Implications for annual effective dose, 14C/12C ratio, and magnetic field effects.

PloS one·2025

相关实验视频

Updated: Sep 14, 2025

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
05:41

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis

Published on: February 6, 2020

9.5K

多摄像头时空深度学习框架,用于在密集的城市环境中实时检测异常行为.

Sai Babu Veesam1, B Tarakeswara Rao2, Zarina Begum3

  • 1School of Computer Science and Engineering, VIT-AP University, Amaravathi, 522241, India. saibabuv@gmail.com.

Scientific reports
|July 23, 2025
PubMed
概括

这项研究引入了一种深度学习框架,用于多摄像头检测异常行为,大大降低了假阳性和计算成本. 它通过改善对未见异常的概括和降低检测延迟来增强实时人群监控.

关键词:
异常检测检测异常检测图表注意力网络的图表.多摄像头监控监控系统强化学习是一种强化学习.时间空间学习学习.

更多相关视频

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

637
Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring
06:32

Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring

Published on: July 14, 2023

1.4K

相关实验视频

Last Updated: Sep 14, 2025

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
05:41

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis

Published on: February 6, 2020

9.5K
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

637
Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring
06:32

Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring

Published on: July 14, 2023

1.4K

科学领域:

  • 计算机视觉 计算机视觉
  • 人工智能的人工智能
  • 监控系统 监控系统

背景情况:

  • 由于人口密度不断增加,城市环境在实时检测异常方面面临挑战.
  • 现有的方法在阻塞,动态场景和计算效率低下方面扎,导致错误的阳性和糟糕的概括.
  • 传统和当前的深度学习模型未能在拥挤的场景中捕捉复杂的社会相互作用和时空依赖.

研究的目的:

  • 提出一种新的深度学习框架,用于多摄像头使用时空信息检测异常行为.
  • 解决现有方法在处理复杂相互作用,计算负载和对未见异常的概括方面存在的局限性.
  • 通过适应性可扩展性和资源配置来增强实时人群监控能力.

主要方法:

  • 多尺度图表注意网络 (MS-GAT) 用于交互意识的异常检测.
  • 基于强化学习的动态摄像头注意力转换器 (RL-DCAT) 用于优化监视焦点和减少计算开销.
  • 空间时空反向对比学习 (STICL) 具有异常记忆,以改善对罕见异常的概括.
  • 基于事件的神经形态编码,使用尖端神经网络进行快速动作分析.
  • 生成性行为合成和代学习的少量适应 (BGS-MFA),用于合成新的异常行为和少量适应.

主要成果:

  • 通过MS-GAT,假阳性病例的减少率高达30%.
  • RL-DCAT 降低了 40% 的计算开销,并增加了 15% 的回忆.
  • STICL提高了未见异常的回忆率25%.
  • 神经形态编码降低了60%的检测延迟.
  • BGS-MFA 提高了异常检测概括率的 35%.
  • 总体框架评估显示,虚假警报减少了40%,计算需求降低了50%,基准数据集的实时效率降低了98%.

结论:

  • 拟议的多方面的深度学习框架有效地解决了当前异常行为检测系统的局限性.
  • 集成MS-GAT,RL-DCAT,STICL,神经形编码和BGS-MFA,为实时多摄像头人群监控提供了一个强大的解决方案.
  • 该框架在准确性,效率和通用性方面取得了显著的改进,为先进的现实应用铺平了道路.