{
  "paper_id": "0cc205dd73191ac635af7021467717ba",
  "paper_title": "Lightweight LiDAR-Based 3D Human Pose Estimation via 2D Depth Images for Autonomous Driving",
  "paper_authors": [
    "Gyu-Yeon Kim",
    "Somi Park",
    "Sunkyung Lee",
    "Bobin Seo",
    "Seon-Han Choi",
    "Sung-Min Park"
  ],
  "publication_date": "2026-03-05",
  "publication_venue": "Sensors",
  "abstract": "Real-world traffic is highly dynamic, with pedestrians exhibiting unpredictable movements. Pedestrians' poses are essential cues for predicting their actions, enabling vehicles to respond proactively and reduce accident risks. In autonomous driving, the distance between vehicles and pedestrians is critical, making 3D human pose estimation crucial. In this context, pedestrian pose estimation has been actively studied, and recently, light detection and ranging (LiDAR) sensors have attracted attention due to their accurate 3D depth information and privacy benefits. However, existing LiDAR-based 3D pose estimation methods mainly process 3D data directly, requiring high computational cost and memory. In this paper, we propose a lightweight LiDAR-based 3D human pose estimation method specifically designed for deployment in autonomous driving systems. Unlike conventional 3D direct processing methods, our approach strategically reduces computational complexity by projecting point clouds into 2D depth images and leveraging a lightweight MoveNet, followed by efficient 3D lifting. Furthermore, we introduce a self-occlusion correction algorithm to improve robustness under side-view and bending poses, where depth-based projections often suffer from distortion. Experimental results on benchmark datasets demonstrate that the proposed method achieves competitive pose estimation accuracy while substantially improving efficiency, highlighting its practicality and scalability for real-time autonomous vehicle applications.",
  "keywords": [
    "autonomous driving",
    "LiDAR",
    "2D depth image",
    "3D human pose estimation",
    "lightweight"
  ],
  "paper_institutions": [
    "Ewha Womans University"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:46:48.822749Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文针对自动驾驶中行人3D姿态估计的计算开销高和自遮挡问题，提出了一种轻量化的基于LiDAR的3D人体姿态估计方法。现有LiDAR方法多直接处理3D点云，依赖PointNet++等复杂结构，计算成本高、内存占用大，不适合实时部署。该方法将点云投影为2D深度图像，利用轻量级MoveNet进行2D关键点估计，再通过深度值进行3D提升，并设计了基于H36M骨架几何对称性和长度先验的自遮挡校正算法。实验采用LiDAR Human26M数据集，包含13个被试、20种动作、传感器距离12-28米，评估指标包括MPJPE、PA-MPJPE、PCK0.5以及参数量和FPS。结果显示提出的方法以仅1.9M参数和440 FPS（GPU）实现PA-MPJPE 108.5mm、PCK0.5 89.6%，参数量较LiDARCap减少约18倍，速度提升超3倍；自遮挡校正使MPJPE和PA-MPJPE分别降低9.2%和11.5%。未来将探索稀疏低分辨率LiDAR下的鲁棒性、方向歧义处理、多人群组泛化和场景级多目标跟踪。该方法适用于资源受限的嵌入式系统，但存在深度图前后歧义导致的姿态翻转失败案例（占0.28%）。核心机制是通过将3D点云转换为2D深度图，再利用深度值直接赋予2D关键点z坐标实现轻量3D重建。\n### 关键要点\n- 现有LiDAR姿态估计多基于PointNet++直接处理3D点云，计算和内存开销大，不适用于实时自动驾驶。\n- 该方法将点云投影为2D深度图像，利用轻量级MoveNet进行2D关键点估计，再结合深度值进行3D提升，简化了流程。\n- 自遮挡校正算法基于H36M骨架，利用左右身对称性和关节长度比例先验，修正侧视和弯曲姿态下的深度失真。\n- 实验使用LiDAR Human26M数据集，数据来自13名被试在户外执行20种动作，距离12-28米，传感器高度5米和7米。\n- 评估指标包括MPJPE、PA-MPJPE、PCK0.5、参数量和FPS，所有方法在同一GPU环境中比较。\n- 提出的方法在GPU上达到440 FPS，参数量仅1.9M，较LiDARCap（34.93M params，138.86 FPS）大幅提升效率。\n- 在CPU（i9-12900K）上达到约11.6 FPS，表明可用于无GPU的低算力环境。\n- 全配置模型获得PA-MPJPE 108.5mm、PCK0.5 89.6%，MPJPE 138.1mm，整体骨架结构保持较好。\n- 躯干和头部区域PA-MPJPE较低（约93mm和89mm），四肢误差较大（手臂158mm、腿120mm），说明核心结构稳健。\n- 距离分析显示中距离（17-22m）误差略低，远距离（22-28m）误差略有增加，但差异不大，表现稳定。\n- 消融实验表明侧遮挡校正降低MPJPE 12mm，弯曲校正效果有限，两者结合达到最佳。\n- 失败案例中，深度图像缺乏纹理导致前后方向歧义，样本中出现翻转姿态，占0.28%，PA-MPJPE高达221mm。",
  "available_templates": [
    "deep_read",
    "simple"
  ]
}