Scholarlink

In large container port yard environments, the presence of numerous repetitive features and certain degenerate scenarios makes it challenging for drones to achieve efficient and reliable SLAM. To address this issue, a drone SLAM method with external normal vector constraints for large container ports is proposed in this paper. First, the IMU and LiDAR data are jointly utilized to extract the normal vectors of the upper surfaces of containers at different layers, and the semantic point clouds of each container layer are partitioned based on container elevation information. Then, a progressive adaptive LiDAR odometry based on semantic point clouds is constructed. This leverages semantic point clouds to adaptively identify degenerate scenarios while optimizing point cloud matching between adjacent frames, thereby enhancing the accuracy and efficiency of the SLAM algorithm. Finally, the effectiveness of the proposed method is verified through testing in Carla simulated scenarios.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

UAV-Based SLAM in Large-Scale Container Port Yards with Outer Normal Vector Constraints

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jie Meng,
  • Zhihang Zuo,
  • Jikai Zhang,
  • Chenghui Wang,
  • Haoyu Cheng,
  • Zhaozheng Hu

摘要

In large container port yard environments, the presence of numerous repetitive features and certain degenerate scenarios makes it challenging for drones to achieve efficient and reliable SLAM. To address this issue, a drone SLAM method with external normal vector constraints for large container ports is proposed in this paper. First, the IMU and LiDAR data are jointly utilized to extract the normal vectors of the upper surfaces of containers at different layers, and the semantic point clouds of each container layer are partitioned based on container elevation information. Then, a progressive adaptive LiDAR odometry based on semantic point clouds is constructed. This leverages semantic point clouds to adaptively identify degenerate scenarios while optimizing point cloud matching between adjacent frames, thereby enhancing the accuracy and efficiency of the SLAM algorithm. Finally, the effectiveness of the proposed method is verified through testing in Carla simulated scenarios.