search

weisongwen / researchTools

useful blogs for research
203 stars 42 forks source link

LiDAR completitions #1

Closed weisongwen closed 4 years ago

weisongwen commented 4 years ago

image

weisongwen commented 4 years ago

Super resolution based map image

weisongwen commented 4 years ago

Input point clouds 16 channels

image

weisongwen commented 4 years ago

ground truth data (64 channels) image

weisongwen commented 4 years ago

Conventional: RPE w.r.t. translation part (m) for delta = 1 (frames) using consecutive pairs (with SE(3) Umeyama alignment)

   max  0.784249
  mean  0.247200
median  0.205328
   min  0.025025
  rmse  0.294806
   sse  2.520408
   std  0.160633

Proposed: RPE w.r.t. translation part (m) for delta = 1 (frames) using consecutive pairs (with SE(3) Umeyama alignment)

   max  1.200751
  mean  0.250122
median  0.151424
   min  0.023755
  rmse  0.371250
   sse  4.134806
   std  0.274346
weisongwen commented 4 years ago

Conclusion:

  1. the density of the maps from SR and the ground truth is comparably the same. In other words, the density is improved compared with the 16 channels data.
  2. the accuracy of the SLAM is pretty similar.
  3. the accuracy of the map still needs to be enhanced. More challenging environments should also be tested.