Scholarlink

This chapter covers how 3D data is represented and processed using voxels, point clouds, and meshes, with methods like PointNet and DGCNN. It discusses early and late fusion strategies for combining sensor data, emphasizing LiDARcamera fusion techniques such as Frustum PointNets and PointPainting to improve object detection. Additionally, feature-level fusion methods like DeepFusion and BEVFusion improve 3D perception by aligning sensor data for more accurate tracking and detection.

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

Robot Perception: 3D Data and Sensor Fusion

  • 章节

  • 发表日期：

    2025年

查看全文
  • Alishba Imran,
  • Keerthana Gopalakrishnan

摘要

This chapter covers how 3D data is represented and processed using voxels, point clouds, and meshes, with methods like PointNet and DGCNN. It discusses early and late fusion strategies for combining sensor data, emphasizing LiDARcamera fusion techniques such as Frustum PointNets and PointPainting to improve object detection. Additionally, feature-level fusion methods like DeepFusion and BEVFusion improve 3D perception by aligning sensor data for more accurate tracking and detection.