What is the task in Distance computation? Selecting the option:
- Enable farthest points removal
- Data: use displayed S.F. (absolute values) as weights
- Model: use displayed S.F. (absolute values) as weights
Without selecting these options, fitting the cloud to the cloud is worse. By selecting these 3 options, the fit of the cloud to the cloud improves a lot, but is there no distortion in the subsequent comparison of the cloud / cloud distance?
Clouds registration (finely registers already, aligned entities)
Re: Clouds registration (finely registers already, aligned entities)
Well, normally you shouldn't use these options (but have you set the 'overlap' parameter correctly? This is the most important parameter).
But to answer your question:
- Enable farthest points removal: as its name implies, at each iteration the farthest points of the to-be-aligned entity are 'removed' and will be ignored during the next iterations
- Data: use displayed S.F. (absolute values) as weights: if the to-be-aligned cloud has a displayed scalar field, its values will be used as weights so that the points with the biggest values have more influence on the registration process than the points with smaller values
- Model: use displayed S.F. (absolute values) as weights : same for the reference cloud
But to answer your question:
- Enable farthest points removal: as its name implies, at each iteration the farthest points of the to-be-aligned entity are 'removed' and will be ignored during the next iterations
- Data: use displayed S.F. (absolute values) as weights: if the to-be-aligned cloud has a displayed scalar field, its values will be used as weights so that the points with the biggest values have more influence on the registration process than the points with smaller values
- Model: use displayed S.F. (absolute values) as weights : same for the reference cloud
Daniel, CloudCompare admin
Re: Clouds registration (finely registers already, aligned entities)
Where is the "overlap" parameter in the options?
Without selecting these options, the result of "automatic" fitting of the cloud to the cloud is RMS = 0.04. By checking the three options we are talking about, the result is RMS = 0.002. When comparing the cloud / cloud distance, we must have the lowest possible result so that the results are as credible as possible.
Fitting runs at 10,000,000 points
Without selecting these options, the result of "automatic" fitting of the cloud to the cloud is RMS = 0.04. By checking the three options we are talking about, the result is RMS = 0.002. When comparing the cloud / cloud distance, we must have the lowest possible result so that the results are as credible as possible.
Fitting runs at 10,000,000 points
Re: Clouds registration (finely registers already, aligned entities)
The overlap parameter is on the main tab, in blue (below the 'RMS difference' field). At least if you use a not too old version.
Mind that with the options you selected (especially the 'remove farthest point' one), the final RMS is computed with less points, as the farthest points are removed at each iteration. So it will always be smaller, but it's biased. You can compute the distances between the two clouds and use 'Edit > Scalar field > Show histogram' to see average distance, as well as the std. deviation, and the RMS in the console. Then you can really compare the two versions...
Mind that with the options you selected (especially the 'remove farthest point' one), the final RMS is computed with less points, as the farthest points are removed at each iteration. So it will always be smaller, but it's biased. You can compute the distances between the two clouds and use 'Edit > Scalar field > Show histogram' to see average distance, as well as the std. deviation, and the RMS in the console. Then you can really compare the two versions...
Daniel, CloudCompare admin