{
  "paper_id": "b60ba8501258b53d7fb944a85b26a91f",
  "paper_title": "On-the-Fly Spatio-Temporal Human Segmentation of 3D Point Cloud Data By Micro-Size LiDAR",
  "paper_authors": [
    "Yuma Okochi",
    "Hamada Rizk",
    "Hirozumi Yamaguchi"
  ],
  "publication_date": "",
  "publication_venue": "2022 18th International Conference on Intelligent Environments (IE)",
  "abstract": "The technology of 3D recognition is evolving rapidly, enabling unprecedented growth of applications towards human-centric intelligent environments. On top of these applications, human segmentation is a key technology towards analyzing and understanding human mobility in those environments. However, existing segmentation techniques rely on deep learning models, which are computationally intensive and data-hungry solutions. This hinders their practical deployment on edge devices in realistic environments. In this paper, we introduce a novel micro-size LiDAR device for understanding human mobility in the surrounding environment. The device is supplied with an on-device lightweight human segmentation technique for the captured 3D point cloud data using density-based clustering. The proposed technique significantly reduces the computational complexity of the clustering algorithm by leveraging the Spatio-temporal relation between consecutive frames. We implemented and evaluated the proposed technique in a real-world environment. The results show that the proposed technique obtains a human segmentation accuracy of 99% with a drastic reduction of the processing time by 66%.",
  "keywords": [
    "3D point cloud",
    "Micro-size LiDAR",
    "segmentation",
    "wearable device",
    "time-series 3D processing"
  ],
  "paper_institutions": [
    "Graduate School of Information Science and Technology, Osaka University",
    "Tanta University"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:47:09.973358Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文针对现有3D点云人体分割依赖大规模LiDAR和高计算量深度学习模型、难以在边缘设备上部署的问题，提出了一种基于微型LiDAR（Hitonavi-μ）的实时人体分割系统。该系统使用改进的DBSCAN密度聚类算法，利用连续帧间的时空关系，仅对新增点云进行聚类，将计算复杂度从O(n²)降至O(m²)，并提取聚类的高度和协方差矩阵等特征，通过SVC分类器区分人体和墙壁。在真实环境中，使用固定在1米高度的Hitonavi-μ采集15m×25m房间内的1776帧点云数据（包含6种人类移动模式），在240帧测试集上评估，与基本DBSCAN相比，处理时间平均减少66%（空环境减少71.1%，高动态场景至少减少60.6%），而分割性能保持高精度（F1-score可达99%）。实验验证了该方法在边缘设备上的实时性和实用性，并为未来在拥挤环境中的扩展提供了方向。\n\n### 关键要点\n- 提出了一种新型微型LiDAR设备Hitonavi-μ，尺寸小、便携，可用于可穿戴场景，实现隐私保护的点云采集。\n- 现有3D分割方法（如PointNet、PointPillars）依赖深度学习，计算密集且数据需求量大，不适合边缘部署。\n- 改进的DBSCAN算法利用连续帧之间的空间关系，仅聚类新出现的点，显著降低计算复杂度。\n- 算法包括bounding_box、update_label、uncontained、include和combine等函数，实现跨帧标签传播和新点聚类合并。\n- 使用SVC分类模型，基于聚类的高度和方差-协方差矩阵六个元素等七个特征进行人体与墙壁的区分。\n- 在真实测试环境中，Hitonavi-μ以10fps帧率固定在1米高度，采集了1776帧点云数据。\n- 测试集包含240帧，覆盖六种人体移动模式（左右通过、接近、离开等），并包含有/无墙壁情况。\n- 与基本DBSCAN相比，改进算法在四种测试场景下处理时间平均减少66%，其中空环境减少71.1%。\n- 改进算法的人体分割F1-score在接近场景达到99.5%，但在通过和离开场景有所下降（79.8%和89.6%），而墙壁分割F1-score保持在92.4%以上。\n- 实验结果验证了实时处理的可行性，平均每帧处理时间33.9ms，满足10fps的要求。\n- 未来工作将研究在拥挤环境中的系统性能，并分析参数变化对系统的影响。"
}