Sensing & Data Preprocessing


LiDAR Preprocessing

High-level Architecture


Input Types


LiDAR Preprocessing

A minimum amount of information is needed to produce correct results. Preprocessing provides sufficient statistics for further algorithms.

  • Remove noisy data from problematic areas
    • Range & angle based filtering
    • Ground filtering
  • Remove redundant data
    • Voxel grids or other downsampling
  • Produce a single consistent representation
    • Static transforms into common frame
    • Fuse into a single pointcloud
    • Ego-motion is required for high speed use cases to correct for slewing

    • Reference

    [1] Sensing Component Design, Link


    Back to top of the page