Scholarlink

This article primarily encompasses different aspects of autonomous indoor navigation for robots. Vision-based camera sensors have recently been supported Simultaneous Localization And Mapping (SLAM) technology, which is being explored for autonomous navigation. However, the capabilities of SLAM are limited to identifying objects from different perspectives. Hence, point cloud information is integrated with the visual feed for robust and precise mapping of the environment. Upon acquiring discrete point clouds, points are stitched and registered, respectively, to reconstruct the scene. Then, using those stitched point clouds, different obstacles within the same environment are registered too. Subsequently KD-Tree (k-dimensional tree) search algorithm is implemented for classifying the point cloud data, which leads to identifying clusters of objects occupied within the environment. In this entire process, an Intel RealSense stereo depth camera coupled with a mobile PC is used for collecting visual feed and concerned 3D point vectors to emulate this integrated concept.

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

Point Cloud Facilitated ORB-SLAM2 for Robust Autonomous Navigation in a Cluttered Indoor Environment

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kuntal Ghosh,
  • Jukka Heikkonen,
  • Rajeev Kanth

摘要

This article primarily encompasses different aspects of autonomous indoor navigation for robots. Vision-based camera sensors have recently been supported Simultaneous Localization And Mapping (SLAM) technology, which is being explored for autonomous navigation. However, the capabilities of SLAM are limited to identifying objects from different perspectives. Hence, point cloud information is integrated with the visual feed for robust and precise mapping of the environment. Upon acquiring discrete point clouds, points are stitched and registered, respectively, to reconstruct the scene. Then, using those stitched point clouds, different obstacles within the same environment are registered too. Subsequently KD-Tree (k-dimensional tree) search algorithm is implemented for classifying the point cloud data, which leads to identifying clusters of objects occupied within the environment. In this entire process, an Intel RealSense stereo depth camera coupled with a mobile PC is used for collecting visual feed and concerned 3D point vectors to emulate this integrated concept.