Postprocess optimization

[EloyAldao]

Dear XW-HKU:

I use to have a low cpu consuming yaml file for data capture (recording a bag file) and a highly optimized for accuracy yaml file for postprocessing the bag recorded in field.
My question, regarding this, is:

• point_filter_num: I understand that this downsamples the pointcloud, grouping the number of points indicated in the parameter. Is it right? If I want the best result should I use value 1?
• max_iteration: this referes to the maximum iterations on ICP adjustment per scan. Is it right? So I should use around 10?
• filter_size_surf: This I think that is the voxelization size. So 0.5 means that every plane of 0.5x0.5 should be used in the adjustment. Is it right?
• filter_size_map: This value I have no idea what means.

If I want the best fitting, even playing the bag at rate 0.5 or 0.1:

• What parameters do you recomend for the best accuracy?
• Should I modify any of the ones hardcoded in the sources?

Thanks in advance

[XW-HKU]

point_filter_num: You can set to 1 of course, the run time will be double i think.

max_iteration: the influence is very tiny, I actually dont recommend to set it to large.

filter_size_surf: I always use same one as filter_size_map but you can try to set this to a different and smaller value.

filter_size_map: This value is the voxel size in the map. for indoor small environments, you may consider to set to 0.2-0.3, but in outdoor large scale scenes, smaller than 0.4 may lead to worse performance.

I think now there are no more hardcode parameters you need modify for better accuracy.

[EloyAldao]

Ok.
Thank you very much!!