{
  "paper_id": "e816461f87b2be06cfebc6b025b3ff24",
  "paper_title": "Real-Time Structural Detection for Indoor Navigation from 3D LiDAR Using Bird's-Eye-View Images",
  "paper_authors": [
    "Guanliang Li",
    "Pedro Espinosa-Angulo",
    "David Perez-Saura",
    "Santiago Tapia-Fernández"
  ],
  "publication_date": "2026-04-21",
  "publication_venue": "unknown",
  "abstract": "Efficient structural perception is essential for mapping and autonomous navigation on resource-constrained robots. Existing 3D methods are computationally prohibitive, while traditional 2D geometric approaches lack robustness. This paper presents a lightweight, real-time framework that projects 3D LiDAR data into 2D Bird's-Eye-View (BEV) images to enable efficient detection of structural elements relevant to mapping and navigation. Within this representation, we systematically evaluate several feature extraction strategies, including classical geometric techniques (Hough Transform, RANSAC, and LSD) and a deep learning detector based on YOLO-OBB. The resulting detections are integrated through a spatiotemporal fusion module that improves stability and robustness across consecutive frames. Experiments conducted on a standard mobile robotic platform highlight clear performance trade-offs. Classical methods such as Hough and LSD provide fast responses but exhibit strong sensitivity to noise, with LSD producing excessive segment fragmentation that leads to system congestion. RANSAC offers improved robustness but fails to meet real-time constraints. In contrast, the YOLO-OBB-based approach achieves the best balance between robustness and computational efficiency, maintaining an end-to-end latency (satisfying 10 Hz operation) while effectively filtering cluttered observations in a low-power single-board computer (SBC) without using GPU acceleration. The main contribution of this work is a computationally efficient BEV-based perception pipeline enabling reliable real-time structural detection from 3D LiDAR on resource-constrained robotic platforms that cannot rely on GPU-intensive processing. The source code and pre-trained models are publicly available.\n",
  "keywords": [
    "3D LiDAR",
    "Bird's-Eye-View (BEV)",
    "Real-time Robotic Perception",
    "Resource-Constrained Robotics",
    "Structural Identification",
    "Structural Mapping",
    "YOLO-OBB"
  ],
  "paper_institutions": [
    "Universidad Politécnica de Madrid"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "deepseek-v4-flash",
  "prompt_template": "simple",
  "extracted_at": "2026-08-18T10:30:16.367040Z",
  "template_tag": "simple",
  "summary": "### 摘要\n\n本文针对资源受限移动机器人上实时结构感知的挑战，提出一种轻量级检测框架：将3D LiDAR点云投影为2D Bird's-Eye-View（BEV）图像，从而在保留垂直墙体几何特征的同时压缩数据量，并基于ROS2模块化架构实现可替换的特征提取与融合流水线。研究背景涉及SLAM与语义结构建图，现有3D语义分割网络计算开销大、依赖GPU，而传统2D几何方法在杂乱室内环境中鲁棒性不足。该框架系统比较了Hough变换、RANSAC、LSD和YOLOv8n-obb四种策略，其中YOLO-OBB通过定向边界框直接回归墙体和柱子的位置、尺寸与角度；后处理阶段采用局部融合与多帧全局融合，核心机制包括基于自定义距离度量 D_ij=max(Δd/τ_d, Δθ/τ_θ, Δo/τ_o) 的DBSCAN空间聚类，以及基于极坐标参数(ρ,α,d1,d2)的卡尔曼滤波和曼哈顿世界优化。训练数据使用合成图像生成器自动标注并注入真实噪声，实验在车库、走廊、实验室和教室走廊四个真实场景中进行，使用Ouster OS2激光雷达，并在无GPU的ThinkBook和Raspberry Pi 5上验证。结果表明，YOLO-OBB在实验室场景取得最高F1分数0.84±0.05，在Raspberry Pi 5上以640×640输入满足10Hz实时要求，内存占用约570MB，距离误差稳定在约6cm、角度误差约1.0度；Hough和LSD响应快但对噪声敏感，LSD过度分割导致系统拥塞，RANSAC鲁棒性较好但无法满足实时性。进一步探索方向包括补充训练集和改进合成数据以解决小尺度场景中薄墙特征的漏检问题；该方法适用于无法依赖GPU加速的低功耗单板计算机等真实部署场景。\n\n### 关键要点\n\n- 论文提出将3D LiDAR点云投影为2D Bird's-Eye-View（BEV）图像的轻量级感知流水线，以在资源受限机器人上实现实时结构检测。\n- 系统基于ROS2模块化设计，包含数据接口、特征检测和特征融合三个功能模块，各算法可替换以进行公平比较。\n- 特征提取对比了Hough变换、RANSAC、LSD和YOLOv8n-obb；其中YOLO-OBB直接回归定向边界框，输出墙体和柱子的几何参数。\n- 点云降维通过高度阈值过滤地面和天花板，并将(x,y)坐标映射为4096×4096像素的BEV占用栅格，最大场景范围约40米。\n- YOLO训练数据由合成图像生成器自动标注，并注入真实噪声块以模拟传感器伪影，从而缩小仿真与真实场景的域差距。\n- 后处理中的局部融合使用自定义距离度量 D_ij=max(Δd/τ_d, Δθ/τ_θ, Δo/τ_o) 进行DBSCAN空间聚类，合并碎片化线段。\n- 全局融合通过时间数据关联、生命周期管理和卡尔曼滤波更新墙体的极坐标参数(ρ,α,d1,d2)，并利用曼哈顿世界优化生成闭合地图。\n- 在车库、走廊、实验室和教室走廊四个场景中，YOLO-OBB在实验室场景取得最高F1分数0.84±0.05，且精度最高，优于几何方法。\n- 在Raspberry Pi 5上，YOLO（640×640）端到端延迟低于100ms，满足10Hz实时要求；FP32 OpenVINO模型内存占用约570MB（7%）。\n- 几何方法在复杂场景中表现不佳：LSD过度分割导致后端拥塞，RANSAC在低内点率时收敛失败，Hough在杂乱环境中产生大量误检。\n- YOLO在距离误差上保持约6cm、角度误差约1.0度的稳定性，而LSD在教室走廊的距离误差超过15cm。\n- 在教室走廊等小尺度场景中，YOLO对薄墙特征存在漏检，作者认为通过补充训练集和改进合成数据可进一步提升泛化能力。"
}