{
  "paper_id": "7ee798d04bb316ed2ff9395ee0962484",
  "paper_title": "Omni-LIVO: Robust RGB-Colored Multi-Camera Visual-Inertial-LiDAR Odometry via Photometric Migration and ESIKF Fusion",
  "paper_authors": [
    "Yinong Cao",
    "Chenyang Zhang",
    "Xin He",
    "Yuwei Chen",
    "Chengyu Pu",
    "Bingtao Wang",
    "Kaile Wu",
    "Shouzheng Zhu",
    "Fei Han",
    "Shijie Liu",
    "Chunlai Li",
    "Jianyu Wang"
  ],
  "publication_date": "",
  "publication_venue": "IEEE Robotics and Automation Letters",
  "abstract": "Wide field-of-view (FoV) LiDAR sensors provide dense geometry across large environments, but existing LiDAR-inertial-visual odometry (LIVO) systems generally rely on a single camera, limiting their ability to fully exploit LiDAR-derived depth for photometric alignment and scene colorization. We present Omni-LIVO, a tightly coupled multi-camera LIVO system that leverages multi-view observations to comprehensively utilize LiDAR geometric information across extended spatial regions. Omni-LIVO introduces a Cross-View direct alignment strategy that maintains photometric consistency across non-overlapping views, and extends the Error-State Iterated Kalman Filter (ESIKEF) with multi-view updates and adaptive covariance. The system is evaluated on public benchmarks and our custom dataset, showing improved accuracy and robustness over state-of-the-art LIVO, LIO, and visual-inertial SLAM baselines.",
  "keywords": [
    "SLAM",
    "multi-camera system",
    "tightly coupled",
    "visual-inertial-LiDAR fusion",
    "direct visual methods",
    "voxel map"
  ],
  "paper_institutions": [],
  "output_language": "zh",
  "provider": "doubao",
  "model": "glm-5.3",
  "prompt_template": "simple",
  "extracted_at": "2026-09-17T09:26:26.999221Z",
  "template_tag": "simple",
  "summary": "### 摘要\n宽视场（FoV）LiDAR传感器可在大范围环境中提供稠密几何信息，但现有LiDAR-惯性-视觉里程计（LIVO）系统普遍依赖单相机，难以充分利用LiDAR衍生深度进行光度对齐与场景着色。在FAST-LIVO2等紧耦合直接法LIVO以及MIMC-VINS、MAVIS等多相机视觉惯性系统的研究背景下，本文提出Omni-LIVO——一个紧耦合多相机LIVO系统，通过跨视图时间迁移（Cross-View temporal migration）直接对齐策略在非重叠视图中保持光度一致性，并以多视图更新与自适应协方差扩展误差状态迭代卡尔曼滤波器（ESIKF）。系统在Hilti SLAM Challenge 2022/2023、Newer College公开基准及自采数据集（LIVOX MID360 LiDAR，360°水平FoV、10Hz；四台1024×768相机交叉配置，手持与机器人平台）上评估，这些数据涵盖建筑工地、室内走廊与地下等多样环境，用以验证精度与鲁棒性。以绝对轨迹误差（RMSE）为核心指标，系统在27条序列中25条取得最低误差，较FAST-LIVO2平均降低30%（最高82%），同时在NVIDIA Jetson AGX Orin嵌入式平台上保持10Hz实时性能（Hilti平均60.3ms、自采序列40.6ms），计算开销仅为1.16–1.21倍。核心机制为跨视图迁移残差 r_mig^{i→j}(p) = τ_i·W_H[I'_i, π_i(p, t_i)] − τ_j·I'_j(π_j(p, t_j))，借助局部平面假设下的单应性矩阵 H_{i→j}(p) = R_{i→j} + t_{i→j}·n_i^T/d_i 在非重叠相机视图间直接扭曲图像块建立光度约束。未来工作将探索在线回环检测与多传感器标定优化；系统适用于机器人与手持建图等实际场景，但需应对非重叠视图无法特征匹配、异构曝光差异与运动模糊等约束。\n\n### 关键要点\n- Omni-LIVO是紧耦合多相机LiDAR-惯性-视觉里程计系统，在FAST-LIVO2基础上扩展多视图光度约束，以充分利用宽视场LiDAR几何信息并扩展空间覆盖。\n- 跨视图时间迁移策略通过迁移指示符M(p, i→j)判定地图点在相机间的转移，并在迁移时动态切换参考图像块（优先当前相机观测或光度误差最小者），保持非重叠90° FoV视图间的光度连续性。\n- 自适应协方差机制依据上一帧平均光度误差线性映射缩放因子（σ_min=100.0、σ_max=1000.0），并经时间平滑（λ=0.3），对高误差相机（如运动模糊、遮挡）降权、对可靠观测升权。\n- 每台相机维护独立的逆曝光时间参数τ_c并在ESIKF中联合优化（雅可比矩阵扩展至R^{m×(6+N_c)}），图像经渐晕归一化处理，无需估计相机间相对曝光因子即可保证跨相机光度一致性。\n- 评估基于Hilti SLAM Challenge 2022/2023、Newer College公开基准及自采数据集（LIVOX MID360与四相机交叉配置，手持与机器人平台），覆盖建筑、走廊、地下等多样环境。\n- 与FAST-LIVO2、FAST-LIVO、R³LIVE、FAST-LIO2和OpenMAVIS五种基线对比，Omni-LIVO在27条序列中25条取得最低绝对轨迹误差（RMSE），并完成全部评测轨迹。\n- 相比FAST-LIVO2平均误差降低30%，Lower Gallery序列提升高达82%（0.006m对0.033m）。\n- 消融实验显示跨视图迁移在大视角变化序列中提升精度（Cupola 2：0.098m对禁用时的0.146m），自适应协方差在光照变化下增强鲁棒性（Lower Gallery 2：0.083m对0.109m）。\n- 相机数量从1台增至3台带来渐进精度增益，几何复杂场景收益最大（Underground hard：0.096m降至0.055m）。\n- 在NVIDIA Jetson AGX Orin（12核ARM 2.2GHz、64GB RAM）上实现10Hz实时性能，Hilti平均60.3ms、自采序列40.6ms，内存分别为2906MB与2359MB，开销为FAST-LIVO2的1.16–1.21倍。\n- RGB着色点云数量显著多于对比方法，basement3场景中着色点数达FAST-LIVO2的3.5倍（33,513,608对9,519,125）。\n- 系统尚未集成在线回环检测，未来将探索在线回环与多传感器标定优化。"
}