Electro-Optic Technology Application, Volume. 39, Issue 6, 35(2024)

Natural Object Recognition Technology Based on Semantic Enhancement Point Cloud Image

JIA Runze1, LI Yuhai1, MA Xiaowen2, SONG Yiheng1, and XU Xinyang1
Author Affiliations
  • 1National Key Laboratory of Electromagnetic Space Security, Tianjin, China
  • 2Xi’an Jiaotong University, Xi’an, China
  • show less

    Aiming at the need for rapid sensing of target terrain in a large range of complex natural object environments in the wild, mainly in mountains and suburbs, an efficient 3D environment terrain segmentation technology is proposed, which combines the point algorithm based on Patch Work, the target point cloud clustering algorithm based on curved-voxel clustering (CVC), the semantic segmentation algorithm based on image and the terrain and the terrain segmentation algorithm based on image and laser fusion. Based on this technology, the advantage of information fusion between laser radar and visible light sensor is used to solve the problem of selecting ground objects in the all-day working state. Though technology development and application integration, the technology scheme proposed can cooperate with the photoelectric system to detect the ground object quickly and efficiently in the complex environment.

    Tools

    Get Citation

    Copy Citation Text

    JIA Runze, LI Yuhai, MA Xiaowen, SONG Yiheng, XU Xinyang. Natural Object Recognition Technology Based on Semantic Enhancement Point Cloud Image[J]. Electro-Optic Technology Application, 2024, 39(6): 35

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category:

    Received: Sep. 12, 2024

    Accepted: Feb. 18, 2025

    Published Online: Feb. 18, 2025

    The Author Email:

    DOI:

    CSTR:32186.14.

    Topics