Hi there,
I'm new to this programme, but I'm super happy with it so far - thanks for making it free :)
I've been looking at cloud-mesh distances and I have a few questions. Any info on any of them would be much appreciated.
1) Is it possible to export the difference output? Other than as an image - it would be great to be able to get at the data.
2) I'm wondering if there is a way to convert from distances into volumes? (I'm looking at erosion so it would be great to be able to calculate the volume of eroded material, but I'm not sure if this is valid or possible.)
3) I'm also wondering what is more valid - a cloud-mesh comparison or a cloud-cloud comparison with quadratic local modeling?
Many thanks in advance for any help / thoughts - all will be gratefully received!!
Cloud-mesh distance
Re: Cloud-mesh distance
Hi,
1) The only way to export the result is to export the cloud (with the distances - i.e. the 'compared' one) as an ASCII file. In such a file you'll have 'X Y Z Dist' on each line. And you can load this kind of file in Matlab or Excel for instance. I believe Some other programs can load this kind of file as well (Paraview, etc.)
2) Not directly. Especially since we use the 'nearest distance' by default and it has no preferred orientation, etc. Some people have tried to use the M3C2 plugin to estimate the volume (search about this on the forum). But it's very approximate.
You can also take a look at the new '2.5D Volume' computation tool.
And the last option is to create a closed mesh of the volume you want to measure with qPoissonRecon (merging the 'before' and 'after' parts of the changing part, and tinkering a lot so as to get the normals pointing in the right direction!). Once you have a closed mesh you can call 'Edit > Mesh > Measure volume'.
3) If you have a mesh, it's always better than a local model. But of course you'd better have a clean mesh ;)
1) The only way to export the result is to export the cloud (with the distances - i.e. the 'compared' one) as an ASCII file. In such a file you'll have 'X Y Z Dist' on each line. And you can load this kind of file in Matlab or Excel for instance. I believe Some other programs can load this kind of file as well (Paraview, etc.)
2) Not directly. Especially since we use the 'nearest distance' by default and it has no preferred orientation, etc. Some people have tried to use the M3C2 plugin to estimate the volume (search about this on the forum). But it's very approximate.
You can also take a look at the new '2.5D Volume' computation tool.
And the last option is to create a closed mesh of the volume you want to measure with qPoissonRecon (merging the 'before' and 'after' parts of the changing part, and tinkering a lot so as to get the normals pointing in the right direction!). Once you have a closed mesh you can call 'Edit > Mesh > Measure volume'.
3) If you have a mesh, it's always better than a local model. But of course you'd better have a clean mesh ;)
Daniel, CloudCompare admin
Re: Cloud-mesh distance
HI there,
I'd like to add a couple of questions about this topic. I am doing a series of measures between several clouds of point compared with the same ground truth (a reference mesh). What I am interested in is the mean and variance of the unsigned distance. I can accomplish this by using the UI, but what about the command line?
1) I couldn't find a way to return the mean and variance from the cmd.
2) I couldn't find a way to select the unsigned distance switch on the cmd.
Are they possible?
Thank you
I'd like to add a couple of questions about this topic. I am doing a series of measures between several clouds of point compared with the same ground truth (a reference mesh). What I am interested in is the mean and variance of the unsigned distance. I can accomplish this by using the UI, but what about the command line?
1) I couldn't find a way to return the mean and variance from the cmd.
2) I couldn't find a way to select the unsigned distance switch on the cmd.
Are they possible?
Thank you
Re: Cloud-mesh distance
Indeed, right now you can't ask for unsigned distances via the command line (mainly because few people are interested in). Beware that the unsigned distances never follow a Normal distribution (therefore the mean and standard deviation doesn't have much meaning).
And there's already a task in the TODO list to output the mean and std. deviation in the command line tool (but it has not been implemented yet).
And there's already a task in the TODO list to output the mean and std. deviation in the command line tool (but it has not been implemented yet).
Daniel, CloudCompare admin
Re: Cloud-mesh distance
Thank you for your answer.
Regarding the need of unsigned distance... maybe I didn't understand correctly how it works.
My goal at the moment is to measure how much a cloud fits to a ground-truth mesh for a bundle adjustment problem. I suppose the signed distance reveals if the point falls inside or outside the mesh, isn't it? If so, from my point of view they're the same kind of error and what I'd need is something like a mean squared error. That's way I went for an unsigned mean. Do you suggest anything better? I am also open to different workflow or softwares..
Thank you again
Regarding the need of unsigned distance... maybe I didn't understand correctly how it works.
My goal at the moment is to measure how much a cloud fits to a ground-truth mesh for a bundle adjustment problem. I suppose the signed distance reveals if the point falls inside or outside the mesh, isn't it? If so, from my point of view they're the same kind of error and what I'd need is something like a mean squared error. That's way I went for an unsigned mean. Do you suggest anything better? I am also open to different workflow or softwares..
Thank you again
Re: Cloud-mesh distance
Indeed, in this case you don't need the signed distances. And the standard deviation is more or less equivalent to the RMS.
Daniel, CloudCompare admin