search

LimHyungTae / patchwork

SOTA fast and robust ground segmentation using 3D point cloud (accepted in RA-L'21 w/ IROS'21)
MIT License
481 stars 75 forks source link

Magic numbers in code #23

Closed james-yoo closed 1 year ago

james-yoo commented 1 year ago

Hi @LimHyungTae, Thanks for releasing the great work! I found that there are magic numbers in the code that can be effected the final extraction result as follow. 1) Could you please explain linearity? https://github.com/LimHyungTae/patchwork/blob/93b493cb57d205dc3564229fd3f9d35977f39ecd/include/patchwork/patchwork.hpp#L460 2) Should I use a value larger than the sensor's mounting height? https://github.com/LimHyungTae/patchwork/blob/93b493cb57d205dc3564229fd3f9d35977f39ecd/include/patchwork/patchwork.hpp#L525 3) What is the meaning of the size of the point cloud? https://github.com/LimHyungTae/patchwork/blob/93b493cb57d205dc3564229fd3f9d35977f39ecd/include/patchwork/patchwork.hpp#L283

james-yoo commented 1 year ago

ping @LimHyungTae,

LimHyungTae commented 1 year ago

True:

  1. If linearity is too large, it means that ground points are likely to be a single channel, which should be avoided when getting average height via ATAT
  2. To reject some undesirable noisy points
  3. Just for reserve mode to prevent inefficient memory copy & paste on the 0-th operation
james-yoo commented 1 year ago

Great. Thanks!