Scholarlink

<p>Traditional vision-based and wireless navigation systems limit visually impaired users due to occlusions, limited camera views, environmental interference, and signal blockage. Continuous tracking in non-overlapping multi-camera networks presents a key challenge in dynamic environments. This paper introduces CamTrack, a novel vision-assisted navigation framework designed to overcome these issues and provide enhanced mobility support for visually impaired individuals. CamTrack employs a novel sensor fusion methodology that unifies existing surveillance camera infrastructure with readily available smartphone sensors and radios, specifically the Inertial Measurement Unit (IMU) and Bluetooth Low Energy (BLE), enabling practical and scalable deployment. The system leverages object detection and deep learning person Re-identification to enable seamless, continuous, and unobtrusive user tracking across multi-camera zones. A key technical contribution is a predictive handover mechanism that utilizes IMU-derived velocity and orientation to anticipate user trajectories, thereby optimizing camera resource allocation, reducing computational overhead, and minimizing latency during transitions. To further enhance robustness, we introduce a camera-handover probability model. Real-world deployment in a smart campus environment validated CamTrack’s effectiveness and practicality in delivering robust, real-time assistive navigation. The analysis of the handover model showed that success rates are lower in sparse camera areas and negatively affected by time spent outside coverage. Quantitative evaluation shows that increasing the number of active cameras from six to nine improves tracking success probability by up to 39% in sparse coverage zones and 13% in dense zones, though at the cost of higher resource utilization. This measurable trade-off highlights how CamTrack enables administrators to balance tracking continuity against deployment cost. It also highlighted the impact of the handover expiry timer on success probability. CamTrack’s use of existing infrastructure offers a scalable and cost-effective solution for smart environments, enhancing accessible navigation.</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

Multi-sensor fusion for user matching and continuous tracking leveraging camera handover in surveillance systems

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月9日

查看全文
  • Sok-Ian Sou,
  • Yi-Cheng Huang,
  • Chiung-Lin Wu

摘要

Traditional vision-based and wireless navigation systems limit visually impaired users due to occlusions, limited camera views, environmental interference, and signal blockage. Continuous tracking in non-overlapping multi-camera networks presents a key challenge in dynamic environments. This paper introduces CamTrack, a novel vision-assisted navigation framework designed to overcome these issues and provide enhanced mobility support for visually impaired individuals. CamTrack employs a novel sensor fusion methodology that unifies existing surveillance camera infrastructure with readily available smartphone sensors and radios, specifically the Inertial Measurement Unit (IMU) and Bluetooth Low Energy (BLE), enabling practical and scalable deployment. The system leverages object detection and deep learning person Re-identification to enable seamless, continuous, and unobtrusive user tracking across multi-camera zones. A key technical contribution is a predictive handover mechanism that utilizes IMU-derived velocity and orientation to anticipate user trajectories, thereby optimizing camera resource allocation, reducing computational overhead, and minimizing latency during transitions. To further enhance robustness, we introduce a camera-handover probability model. Real-world deployment in a smart campus environment validated CamTrack’s effectiveness and practicality in delivering robust, real-time assistive navigation. The analysis of the handover model showed that success rates are lower in sparse camera areas and negatively affected by time spent outside coverage. Quantitative evaluation shows that increasing the number of active cameras from six to nine improves tracking success probability by up to 39% in sparse coverage zones and 13% in dense zones, though at the cost of higher resource utilization. This measurable trade-off highlights how CamTrack enables administrators to balance tracking continuity against deployment cost. It also highlighted the impact of the handover expiry timer on success probability. CamTrack’s use of existing infrastructure offers a scalable and cost-effective solution for smart environments, enhancing accessible navigation.