{
  "paper_id": "c89e6df34899588865563de94a6602e1",
  "paper_title": "QuadricsReg: Large-Scale Point Cloud Registration Using Semantic Quadric Primitives",
  "paper_authors": [
    "Ji Wu",
    "Huai Yu",
    "Shu Han",
    "Ximeng Cai",
    "Mingfeng Wang",
    "Wen Yang",
    "Gui-Song Xia"
  ],
  "publication_date": "",
  "publication_venue": "IEEE Transactions on Robotics",
  "abstract": "Designing an effective and scalable scene primitive representation is fundamental for large-scale point cloud registration. Existing studies that predominantly rely on dense point clouds or single-type geometric primitives struggle to scale to scenes characterized by massive data volume, structural diversity, and wide viewpoint variations. To address these, this article introduces QuadricsReg, a novel point cloud registration framework based on quadric primitives for large-scale environments. We compactly model diverse scene structures within a unified semantic quadric formulation, achieving high compression while preserving geometric richness and discriminability. This representation enables efficient quadric matching initialization via intrinsic similarity and robust correspondence pruning by maximizing geometric consistency in a multilevel graph, ensuring reliable associations even under large viewpoint variations. Furthermore, we design a factor graph based on degeneracy-aware quadric residual to estimate the transformation, ensuring accurate alignment in heterogeneous scenes. We evaluate QuadricsReg on 5 public datasets, where its exceptional registration performance with low overhead demonstrates strong scalability for large-scale scenarios. With a compact representation of ~29.5KB/scan on KITTI, nearly 100% registration success rate is achieved for point cloud pairs within 10m. Real-world testing on the self-collected dataset further validates its robustness and generalization ability across different LiDAR sensors and robot platforms.",
  "keywords": [
    "Geometric primitives",
    "localization and mapping",
    "point cloud registration",
    "quadric representation"
  ],
  "paper_institutions": [
    "School of Computer Science, Wuhan University",
    "School of Electronic Information, Wuhan University",
    "School of Artificial Intelligence, Wuhan University"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "glm-5.3",
  "prompt_template": "simple",
  "extracted_at": "2026-09-17T09:39:22.799546Z",
  "template_tag": "simple",
  "summary": "### 摘要\n\n大规模点云配准是机器人SLAM、大规模建图与自主导航的基础任务，但现有方法多依赖稠密点云或单一类型几何基元（关键点、线、平面、圆柱、椭球等），难以应对海量数据量、结构多样性与大视角变化带来的挑战。QuadricsReg提出基于语义二次曲面的全局配准框架：利用仅10个参数的二次曲面隐式方程 x^T Q x = 0 统一表示17类常见几何形状，将场景抽象为包含语义、位姿、尺度与退化属性的21参数紧凑表示（KITTI单帧约12万点压缩至约100个二次曲面、约29.5KB）；通过语义类型内的尺度相似度初始化对应关系，并构建多级兼容图以最大团搜索剔除外点；进而以SVD初始化6-DoF变换，在因子图中利用退化感知的二次曲面残差（统一point-to-point/line/plane度量）进行优化。方法在KITTI、KITTI-360、Apollo、Waymo、nuScenes五个公开数据集及自采Hetero-Reg数据集（UGV/UAV/手持平台，Velodyne-16、Livox Mid-360、Livox Avia三种LiDAR）上评估，采用内点率(IR)、对应召回率(CR)、配准成功率(SR)、相对平移/旋转误差(RTE/RRE)等指标：10m内点云对配准成功率接近100%，硬级别(20-30m)下仍显著优于FCGF、DGR、PointDSC、RANSAC、TEASER++、Quatro、Segregator、G3Reg等基线；存储开销不足点云表示的三十分之一，后端以默认约2Hz、快速模式5Hz运行。真实场景验证涵盖LIO-SAM回环闭合、跨平台多会话建图及森林/走廊等退化场景。未来方向包括融合图像纹理增强提取、语义与几何的自适应集成建模以及动态实体滤波。其核心机制为二次曲面隐式方程 x^T Q x = 0 及其分解 Q = P_q^{-T} C P_q^{-1}，通过退化指示子 I_{s,R,t} 在优化中解耦有效几何约束与虚假自由度。\n\n### 关键要点\n\n- 提出QuadricsReg框架，以语义二次曲面统一表示异构场景结构：仅10参数即可建模17类几何基元（点、线、平面、球、圆柱等），兼顾紧凑性、表达力与判别性。\n- 场景表示采用21参数结构（语义标签、二次曲面系数、完整尺度、四元数旋转、中心），典型KITTI扫描（约12万点）压缩至约100个二次曲面，内存占用相当于约700个点。\n- 二次曲面提取分为关键结构（地面、平面、线，基于TRAVEL与G3Reg流程）与关键对象（PVKD语义引导或几何聚类两种互补模式），并通过元素-曲面映射表约束拟合类型。\n- 拟合采用学习法与统计法并行策略：QuadricsNet对噪声与部分观测鲁棒，统计估计保证分布外有效几何，依据点到曲面残差自适应选择最优结果。\n- 匹配阶段在共享语义类内基于尺度相似度建立初始对应（每曲面保留top-K_s候选），并通过四级阈值[0.2, 0.4, 0.6, 0.8]的兼容图与渐进最大团搜索实现多级外点剔除。\n- 变换估计先以SVD基于内点中心初始化，再在因子图中以退化感知二次曲面残差优化：旋转项仅惩罚非退化主轴失配，平移项将误差投影至有效子空间，自然退化为point-to-point/line/plane度量。\n- 针对非中心曲面（平面、线、圆柱）补充采样点以恢复约束秩并提供空间杠杆臂，避免中心相交但边界错位时残差消失的问题。\n- 评估覆盖五个公开数据集（KITTI、KITTI-360、Apollo、Waymo、nuScenes）的回环与里程计场景，以及自采Hetero-Reg数据集的SLAM、多会话建图与挑战场景。\n- KITTI-LC上10m内点云对配准成功率接近100%，硬级别（20-30m）下成功率（如序列05达87.5%）显著优于学习法、手工法与高层表示基线。\n- 效率方面，每扫描存储约29.5KB（不足点云表示的三十分之一），后端运行频率默认约2Hz、快速变体5Hz，满足回环闭合需求。\n- 真实场景验证包括：集成于LIO-SAM消除z轴漂移与重访区域地图不一致、跨UGV/UAV/手持平台与三种LiDAR的多会话地图合并、森林与走廊等稀疏退化场景下的鲁棒配准。\n- 鲁棒性分析表明：非结构化森林中IR显著下降但SR保持稳定；语义标签遮蔽50%时易/中级别成功率仍保持高位；局限在于低重叠下二次曲面模型不一致及动态物体干扰，未来将融合图像纹理并估计运动状态以滤除动态实体。"
}