{
  "paper_id": "bd50599d67305ec629d7515dc358dc82",
  "paper_title": "LTM: A Fast and Error-sensitive Point Cloud Clustering Algorithm for LiDAR Panoptic Segmentation",
  "paper_authors": [
    "Peng Wu",
    "Zhengxi Hu",
    "Qianyi Zhang",
    "Jinzheng Guang",
    "Jingtai Liu"
  ],
  "publication_date": "",
  "publication_venue": "2024 9th International Conference on Robotics and Automation Engineering (ICRAE)",
  "abstract": "Point cloud clustering is a crucial approach to understand and analyze geometric information for both robotic systems and autonomous driving. To meet real-time requirements, several existing point cloud clustering algorithms are based on Connected Component Labeling. It means that classification decisions are made locally based on the geometric relationships among neighboring points. However, the existing neighborhood connection criterion lacks robustness. This is because the difference in resolution between point clouds in the horizontal and vertical directions and various depth affect the distance between adjacent points, thereby affecting connectivity determination. In this paper, we propose a novel clustering algorithm LTM (includes Linearization, Trunk extraction and Merge). It processes the horizontal and vertical directions separately in two steps and innovatively utilizes three points to determine neighborhood connectivity. Thus, it can reduce the impact of these constraints on the effectiveness of semantic segmentation models and perform a panoptic segmentation task on the SemanticKITTI. Experimental results demonstrate that LTM achieved the best results among all published methods of the same category. Furthermore, the single-frame processing time for clustering is within 10 milliseconds.",
  "keywords": [
    "LiDAR Panoptic Segmentation",
    "Point Cloud Clustering",
    "Connected Component Labeling"
  ],
  "paper_institutions": [
    "Robotics and Automatic Information System, Nankai University",
    "Tianjin Key Laboratory of Intelligent Robotics",
    "TBI center, Nankai University"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:47:30.587980Z",
  "template_tag": "simple",
  "summary": "### 摘要\n\n本文提出了一种面向LiDAR全景分割的快速且对错误敏感的点云聚类算法LTM（包含线性化、主干提取与合并三个步骤）。现有基于连通组件标记（CCL）的聚类算法通常仅利用两点间的几何关系判断邻接性，但LiDAR点云在水平与垂直方向分辨率不同、深度变化以及入射角差异等因素会导致相邻点间距不稳定，从而影响连通性判定的鲁棒性。LTM通过创新性地引入三点共线约束作为新的邻接判定准则，分别在水平方向和垂直方向处理点云，以降低深度和入射角的影响。在SemanticKITTI数据集上的实验表明，LTM在同类基于CCL的聚类方法中取得了最佳性能，且单帧聚类处理时间在10毫秒以内。该方法可结合现有的语义分割网络（如Cylinder3D和2DPass）完成全景分割任务，在密集交通场景中也能准确区分邻近的同类实例。进一步地，LTM可应用于机器人和自动驾驶中的感知、定位与地图构建等实际场景，其核心机制在于通过三点共线性条件在水平方向提取线性点集，再根据方位角重叠度和鸟瞰图中心点距离提取实例主干，最后利用包围框IoU合并主干并聚类剩余孤立点，从而在保证线性的时间复杂度内实现高精度的实例分割。\n\n### 关键要点\n- LTM算法包含线性化、主干提取和合并三个阶段，整体时间复杂度与点云数量呈线性关系。\n- 创新地利用三点共线性作为邻接判定准则，替代传统的两点距离或角度阈值，减少误判。\n- 水平方向通过相邻三点的深度关系计算偏差δ=|d1+d3-2d1d3/d2|，与阈值比较判断是否属于同一实例。\n- 垂直方向通过计算线性点集之间方位角重叠比率和鸟瞰图中心点距离来提取实例主干。\n- 合并阶段使用包围框交并比（IoU）来聚合多个主干，并利用语义分割先验施加尺寸约束。\n- 在SemanticKITTI验证集（序列08）上，Cylinder3D+LTM达到PQ 59.2，优于Cylinder3D+SLR（56.8）和Cylinder3D+DMC（57.1）。\n- 在SemanticKITTI测试集（序列11-21）上，2DPass+LTM达到PQ 58.7，显著高于2DPass+SLR的55.1和2DPass+DMC的55.9。\n- 与基于两点准则的SLR和Depth Cluster相比，LTM的真阴性数更高、假阳性更低，显著误差更少。\n- 单帧聚类平均运行时间约为10毫秒，满足实时应用需求。\n- LTM能准确分割距离很近的车辆等同类实例，而SLR在类似场景下容易因少量误判导致聚类失败。\n- 研究还探讨了四点准则的局限性，表明三点准则在平衡约束严格性和结构完整性方面更优。\n- 该方法可结合不同语义分割网络，随着语义分割精度提升，混合方案有望达到更高性能。"
}