{
  "paper_id": "4be20029121298d9525d03e2f30de3b1",
  "paper_title": "3D Human Pose and Shape Estimation from LiDAR Point Clouds: A Review",
  "paper_authors": [
    "Salma Galaaoui",
    "Eduardo Valle",
    "David Picard",
    "Nermin Samet"
  ],
  "publication_date": "",
  "publication_venue": "unknown",
  "abstract": "In this paper, we present a comprehensive review of 3D human pose estimation and human mesh recovery from in-the-wild LiDAR point clouds. We compare existing approaches across several key dimensions, and propose a structured taxonomy to classify these methods. Following this taxonomy, we analyze each method's strengths, limitations, and design choices. In addition, (i) we perform a quantitative comparison of the three most widely used datasets, detailing their characteristics; (ii) we compile unified definitions of all evaluation metrics; and (iii) we establish benchmark tables for both tasks on these datasets to enable fair comparisons and promote progress in the field. We also outline open challenges and research directions critical for advancing LiDAR-based 3D human understanding. Moreover, we maintain an accompanying webpage that organizes papers according to our taxonomy and continuously update it with new studies: <https://github.com/valeoi/3D-Human-Pose-Shape-Estimation-from-LiDAR>",
  "keywords": [
    "3D human mesh recovery",
    "3D human pose estimation",
    "3D human pose estimation from LiDAR",
    "3D human recovery from LiDAR",
    "human pose estimation benchmarks",
    "human pose estimation metrics",
    "human mesh recovery benchmarks",
    "LiDAR point clouds",
    "LiDAR human pose estimation datasets",
    "point cloud-based human modeling",
    "review",
    "survey"
  ],
  "paper_institutions": [],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:46:32.394945Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文系统综述了基于野外LiDAR点云的三维人体姿态估计（3D HPE）和人体网格恢复（HMR）方法。针对LiDAR点云稀疏、不规则、噪声大及遮挡等挑战，文章提出了一个结构化分类法，按照监督程度（监督、弱监督、无监督）和方法架构（基于Transformer、稀疏卷积等）对现有方法进行归类，并逐一分析其设计选择、优缺点。在数据集方面，定量比较了Waymo Open Dataset、SLOPER4D和Human-M3三个最常用数据集，给出了外在与内在特性统计，并统一了所有评估指标的定义。文章还建立了在三个数据集上的HPE和HMR基准表，汇总了各方法的性能，并指出了开放挑战和未来方向，如数据稀缺、相机参数依赖、传感器域适应等。文中核心公式为三维姿态多样性度量：dist(Ĵ,J^T) = (1/N_jts) Σ ||Ĵ_i - J_i^T||_2^2，用于衡量不同数据集中人体姿态的多样程度。该综述覆盖2019至2025年的32项研究，并附带持续更新的GitHub资源库，为LiDAR-based人体感知提供了全面的参考资料。\n### 关键要点\n- LiDAR点云具有内在3D几何信息、精确测距、对光照鲁棒、保护隐私等优势，但存在稀疏、不规则、扫描图案依赖、传感器噪声和遮挡等挑战。\n- 文章回顾了3D HPE从图像/视频到LiDAR的发展脉络，并将传感器分为相机、雷达和LiDAR三类，重点介绍了LiDAR的扫描类型（NRS和RMB）及技术。\n- 提出了一种系统分类法，按监督级别（监督、弱监督、无监督）和方法架构（Transformer、稀疏卷积、PointNet等）组织32种现有方法。\n- 监督方法如LiPose、LPFormer、VoxelKP等通过结构化投影、体素化、密度感知变换器等方式应对点云稀疏性。\n- 弱监督方法多依赖RGB-LiDAR融合，利用2D关键点作为监督信号，例如WS-HPE、WS-Fusion、LiCamPose等，减少对3D标注的依赖。\n- 无监督方法GC-KPL利用合成数据预训练和几何约束（对称损失、点-肢体距离等）在无标签数据上学习人体结构。\n- HMR方法如LiDARCap、LiveHPS、NE-3D-HPE等采用时序建模、逆运动学求解和SMPL回归，从稀疏点云重建完整网格。\n- 数据集方面，Waymo Open Dataset覆盖75米范围、64线LiDAR；SLOPER4D采用128线、360°FOV，人均点云密度更高；Human-M3采用NRS扫描，融合四视角点云。\n- 评估指标包括MPJPE、PA-MPJPE、PCK、OKS、MPVPE、MPERE、Ang Err等，文章统一了其数学定义，并标注了方向（↓/↑）。\n- 基准结果显示，在WOD上LPFormer达到PEM 0.1524，DAPT在SLOPER4D上MPJPE 2.80cm，LiveHPS++在SLOPER4D上MPJPE 7.70cm且加速误差434cm/s²，表现领先。\n- 未来方向包括减少对辅助模态的依赖、利用时间信息、生成更真实的合成数据、开发弱监督HMR方法、消除相机校准依赖以及跨传感器域自适应。\n- 文章维护了GitHub资源库，持续更新按分类法组织的论文和数据集统计代码，便于研究者比较和复现。",
  "available_templates": [
    "deep_read",
    "simple"
  ]
}