{
  "paper_id": "9b38a6b20f4d38463cc9f84861bc750a",
  "paper_title": "Recognition of Human Characteristics Using Multiple Mobile Robots With 3D LiDARs",
  "paper_authors": [
    "Koki Makita",
    "Dražen Bršić",
    "Takayuki Kanda"
  ],
  "publication_date": "",
  "publication_venue": "2021 IEEE/SICE International Symposium on System Integration (SII)",
  "abstract": "Mobile robots are gradually entering our human spaces. Apart from being able to accurately localize themselves, in populated environments robots also need to detect humans and recognize their characteristics. In this work we focused on using 3D LiDARs for sensing, and in particular on the question whether they can be used to classify characteristics of people around the robot. Moreover, since using sensors from multiple robots is expected to give more accurate recognition, we compared several ways how to combine multiple 3D LiDARs. We evaluated these methods using simulator data as well as actual real-world data. For the real-world data we created a novel dataset taken with multiple robots equipped with 3D LiDARs in a shopping center, which also includes manually labeled characteristics of the detected pedestrians. The results show that combining several 3D LiDARs makes the recognition more accurate. However, we were not able to achieve satisfactory accuracy with real-world data.",
  "keywords": [
    "3D LiDAR",
    "human characteristics recognition",
    "multiple mobile robots",
    "data fusion",
    "deep fusion",
    "score fusion"
  ],
  "paper_institutions": [],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:47:10.915136Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文旨在解决利用多台移动机器人上的3D LiDAR识别行人特征（如性别、是否背包）的问题。已有研究主要将3D LiDAR用于定位和检测，对行人特征识别研究较少，且多传感器融合的方式尚不明确。我们提出了基于卷积神经网络的方法，比较了单传感器、得分融合（对每个传感器单独分类后取最大值）和深度融合（在网络中间层使用视图池化进行特征融合）三种方式。在仿真环境（MORSE模拟器，6个机器人，98个人体模型）和真实环境（3台机器人在大阪ATC购物中心采集的公开数据集）中进行了评估，识别任务是背包和性别分类。仿真结果表明多传感器融合显著提升准确率，其中得分融合最佳（背包97.1%，性别97.8%）；真实数据上融合也优于单传感器，但准确率仅约60%，远低于仿真结果。核心机制为：得分融合对每个传感器的分类得分取元素级最大值，深度融合在共享卷积层的特征图上进行视图池化（同样取最大值），以适应可变的传感器数量。未来的改进方向包括扩大数据集、融合多时刻信息以及提高真实场景下的鲁棒性。\n\n### 关键要点\n- 3D LiDAR对光照变化鲁棒且保护隐私，但垂直分辨率低导致点云稀疏，是行人特征识别的主要挑战。\n- 本文首次创建了由三台移动机器人携带3D LiDAR在真实商场中采集的行人数据集，包含53,510帧人工标注数据。\n- 融合方法需适应不同时刻可见行人的LiDAR数量变化，视图池化层通过元素级最大值操作有效应对此问题。\n- 仿真数据上得分融合的准确率（背包97.1%、性别97.8%）高于深度融合（89.7%、95.7%）和单传感器（73.5%、75.1%）。\n- 真实数据上深度融合与得分融合准确率接近（约61%），均优于单传感器（约57-59%），但总体水平不足以实际部署。\n- 真实识别率低可能源于仿真人体模型过于简单、真实行人外观多样性高、传感器多位于同侧且距离远等原因。\n- 网络训练使用10折交叉验证和早期停止（最大20个epoch），确保评估公平。\n- 传感器输入为30×40像素的深度图像，由3D点云转换而来，该尺寸平衡了参数数量和信息损失。\n- 真实数据中行人点云提取需要先基于预建3D地图定位机器人，将不匹配地图的激光点聚类为行人，并用卡尔曼滤波跟踪。\n- 未来工作可扩大真实数据集规模，并融合多个时刻的识别结果，后者已有研究证明能显著提高准确率。"
}