{
  "paper_id": "5b76c5afb87d3608392260e2c483d3ba",
  "paper_title": "LiDAR Point Cloud Feature Learning with One-Class SVM for Efficient Human Recognition in Indoor Environment",
  "paper_authors": [
    "Hadil Eltalif",
    "Charles Gouin-Vallerand",
    "Yacine Yaddaden"
  ],
  "publication_date": "",
  "publication_venue": "2026 22nd International Conference on Intelligent Environments (IE)",
  "abstract": "Accurate person detection is a key enabler for intelligent monitoring in smart home environments, supporting applications such as assisted living, activity recognition, and adaptive automation. 3D LiDAR sensors are particularly well suited for this task, as they provide precise geometric information while preserving privacy. However, robust indoor detection remains challenging due to cluttered interiors, occlusions, diverse human activities, and the limited availability of annotated datasets. We propose a novel LiDAR-based person detection approach that leverages cross-dataset transfer between deep feature learning and One-Class classification. Deep geometric features are first learned from the L-CAS 3D People Dataset and transferred to the SZTAKI LiDAR Gait-and-Activity dataset, where they are exploited by a lightweight One-Class SVM trained only on the “person” class. This enables the detection of non-person anomalies without requiring extensive negative samples. We report multiclass classification results on the L-CAS dataset, achieving over 98% accuracy. On the more challenging SZTAKI-LGA dataset, the proposed method reaches up to 96% accuracy and 96% F1-score. Finally, the full pipeline is validated on a small in-house dataset collected in our laboratory, highlighting its practical feasibility in a realistic indoor setting and providing a foundation for future extensions toward multi-person tracking and advanced smart-environment applications.",
  "keywords": [
    "3D LiDAR",
    "Person Detection",
    "Deep Learning",
    "One-Class classification",
    "Smart Indoor Environments"
  ],
  "paper_institutions": [
    "Université de Sherbrooke",
    "Université de Sherbrooke",
    "Université du Québec à Rimouski"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:46:24.680245Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文针对智能室内环境中准确且保护隐私的人员检测问题，提出了一种结合深度几何特征学习与单类支持向量机（OC-SVM）的LiDAR人员检测方法。3D LiDAR传感器能提供精确几何信息且保护隐私，但室内环境存在杂乱、遮挡、活动多样和标注数据稀缺等挑战。现有研究多面向户外自动驾驶，室内场景的跨环境泛化研究不足。方法上，先从L-CAS数据集学习深度几何特征（PointNet、PointNet++ SSG/MSG、PointNet-Attention），再在SZTAKI-LGA数据集上仅用“人”类样本训练轻量级OC-SVM，以检测非人异常而无需大量负样本。在L-CAS上实现超过98%的多类分类精度；在SZTAKI-LGA上达到96%的准确率和F1分数；最后用自制实验室数据（VLP-32C）验证了可行性。结果表明，即使训练数据有限，该两阶段方法在室内现实条件下也能提供稳健、可靠且计算开销低的检测，为未来多目标跟踪和高级智能环境应用奠定基础。核心机制是通过深度编码器提取1024维嵌入，再以OC-SVM学习人形超球边界，决策函数为$f(\\mathbf{x})=\\mathrm{sign}(\\langle\\mathbf{w},\\phi(\\mathbf{x})\\rangle+b)$。\n\n### 关键要点\n- 提出混合检测框架：深度LiDAR特征学习 + 单类SVM，用于隐私保护的室内人员检测。\n- 深度特征在L-CAS上预训练，转移至SZTAKI-LGA，并用OC-SVM实现仅正样本训练。\n- 预处理包括点云归一化、固定点数重采样（256/512/1024）及数据增强（上下分割、偏航旋转）。\n- 比较三种深度架构：PointNet、PointNet++（SSG/MSG）和PointNet-Attention；PointNet++ MSG在L-CAS上最佳（256点，精度99%）。\n- 在SZTAKI-LGA上，PointNet++ SSG和MSG（1024点）均达96%准确率和F1分数。\n- 在自制实验室数据集上，PointNet++ MSG（1024点）达98%准确率、99%精度和97%召回率。\n- 直接跨传感器转移（HDL-32E到VLP-32C）性能下降（最优76%），重新训练OC-SVM后显著改善。\n- OC-SVM只使用正向样本，验证集仅用于调优$\\nu$和决策阈值，确保无偏评估。\n- 在个人工作站上，预处理几十毫秒、特征提取几百毫秒、OC-SVM预测近乎瞬时（0.064毫秒），延迟可接受。\n- L-CAS上，与以往仅用手工特征的方法（精度约46%-60%）相比，本方法精度显著提升（99%）。\n- SZTAKI-LGA没有现有“人vs非人”基准，表明室内LiDAR检测研究尚不充分，本研究填补了空白。"
}