Scholarlink

<p>The rapid proliferation of Internet of Things (IoT) technologies has intensified the demand for intelligent and reliable visual surveillance systems capable of real-time human identification. Among various surveillance tasks, face recognition remains a critical yet challenging component due to variations in pose, illumination, occlusion, and motion inherent in video streams. While existing face recognition approaches predominantly focus on still images and fixed-size inputs, such assumptions significantly limit robustness and scalability in practical IoT surveillance environments. To address these challenges, this work presents a video-based human identification framework designed for IoT surveillance networks that integrates detection, tracking, and recognition in a unified pipeline. The proposed approach employs a motion-aware face detection and tracking mechanism to ensure temporal consistency across video frames, followed by discriminative feature extraction and embedding for identity representation. A learning-driven recognition strategy is adopted to enhance generalization across unconstrained scenarios. Extensive experimental evaluation on benchmark video face datasets demonstrates improved recognition accuracy, robustness under dynamic conditions, and suitability for real-time IoT surveillance applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Learning Based Human Identification Using IoT Surveillance Network Systems

  • 期刊论文

  • 发表日期：

    2026年4月6日

查看全文
  • Srikanth Bethu,
  • Srikanth Lakumarapu,
  • Shabana Mahammad

摘要

The rapid proliferation of Internet of Things (IoT) technologies has intensified the demand for intelligent and reliable visual surveillance systems capable of real-time human identification. Among various surveillance tasks, face recognition remains a critical yet challenging component due to variations in pose, illumination, occlusion, and motion inherent in video streams. While existing face recognition approaches predominantly focus on still images and fixed-size inputs, such assumptions significantly limit robustness and scalability in practical IoT surveillance environments. To address these challenges, this work presents a video-based human identification framework designed for IoT surveillance networks that integrates detection, tracking, and recognition in a unified pipeline. The proposed approach employs a motion-aware face detection and tracking mechanism to ensure temporal consistency across video frames, followed by discriminative feature extraction and embedding for identity representation. A learning-driven recognition strategy is adopted to enhance generalization across unconstrained scenarios. Extensive experimental evaluation on benchmark video face datasets demonstrates improved recognition accuracy, robustness under dynamic conditions, and suitability for real-time IoT surveillance applications.