Could this software help me?

[Kooder]

First of all, please excuse my bad english.

I have to compare two point clouds of person faces. One of them reconstructed with one camera (kinect style) and the other with three cameras (aligned with registration techniques). But I need to compare them not in a visual way, I need some kind of table data with, for example, precision and recall values, or some analytic data which helps me to prove my 3-cam cloud is better than my 1-cam cloud (Benchmark tests?).

I found this awesome software but I'm not sure if it can help me.

Thank you and sorry for my english again!

[daniel]

Well, it sounds to be the case.

First CC will let you compute distances between the two points clouds (warning, if the 'reference' cloud density is too low you might have to use 'local modeling' strategies or even mesh the reference cloud - see http://www.cloudcompare.org/doc/wiki/in ... omputation).

Then from this distance set (one value per point of the compared cloud) you can get simple statistics (min, max, mean and standard deviation). But you can also export the values to Excel or Matlab for instance if you need.

And depending on how the clouds are generated, you may need first to register them (see http://www.cloudcompare.org/doc/wiki/in ... gistration).

And more generally speaking, if you really need to say 'this cloud is better than this cloud' you should compare both clouds to a 'ground-truth' reference (i.e. a third cloud or mesh acquired with a much more accurate and reliable scanner).

[Kooder]

Thank you for your quick answer!

So... If i don't have a third high-quality cloud to compare, it's difficult to prove one cloud is better than the other, right?

Thank you again!

[daniel]

Yes. To say that something is better than something else, you always have to compare both to a (gold) standard. That's the same with currencies ;)

As an alternative the standard could also be something measurable (density, roughness, etc.). But I doubt it will be sufficient here...

[Nautarcher]

Hello Daniel

I am interested in performing a comparison much the same way - both visually and empirically, so I couldn't help but notice that you said you could export those point cloud values to either Excel or Matlab. What specific steps would I need to take to do that?

Also, I noticed the Animation Plug-in on the tool bar. Will that now allow you to perform a 'fly-through' animation of the cloud(s)?

Thanks ahead of time!

Best Regards
Mike

[daniel]

Indeed, once you have computed the distances (either C2C or C2M - see http://www.cloudcompare.org/doc/wiki/in ... omputation) you can export either all the distances values (by exporting the cloud as an ASCII file) or only the histogram (via the 'Histogram' tool).

And indeed the qAnimation plugin can generate fly through videos. See http://www.cloudcompare.org/doc/wiki/in ... n_(plugin).