Comparing two different objects

[pochihlee]

Hi

I want to compare two human ribs (Fig 1) reconstructed by two different software, both of them are STL file. In Fig 1, left rib is taken as reference object. I do C2C (Fig 2) and C2M (Fig 3) comparison, and the unit is meter. Both of the mean (0.0962cm for C2C, 0.034cm for C2M) and the standard deviation (0.0636ccm for C2C, 0.1102cm for C2M) are very small. In your opinion, do you have any suggest to compare these two ribs by CloudCompare? Should I use C2C or C2M or M3C2 (which I think I can not fully understand the method after seeing the paper) to compare these two ribs? Also, how can I explain the result? By the way, I saw paper using "point to surface distances" to compare two reconstruction objects, does it mean I should use C2M to compare two reconstruction objects?

Thank you very much.

[daniel]

If you have meshes, then you should stick with C2M distances. You should either check the "symmetrical" option of the color scale (see the second tab of the SF settings in the cloud properties) or define you own color scale (with the Color Scale Manager).

You'll see more clearly the "intersecting" parts, and the parts which are below or above the reference mesh.

One big question is always how both shapes were registered / aligned before comparing them?

[pochihlee]

For the C2C, I subsample the mesh for both of the mesh.

For the C2M, I subsample the compared cloud. Before comparing, I use the auto align tool to align the mesh (reference rib) and cloud (compared rib), and I set the RMS difference to 1e-10. Would it be ok?

For the C2M, I show the reference rib and the compared rib simultaneously in Fig 1 (frontal view) and Fig 2 (side view). I can clearly identify the part above or below the reference mesh. As you can see in Fig 2, the scalar field shows the tail in red color, but the tail is actually below the reference mesh. Is there anything wrong?

To be clearer, I show the bottom view of the ribs in Fig 3 and I also set the scalar filed to only 4 steps. It seems that CloudCompare thinks the compared rib is above the reference rib (it shows the positive value in SF), but the compared rib is actually below the reference rib.

[orinn]

Below in this scenario, means inside or outside of the other mesh. The tail is outside the mesh that is why it is in Red. The comparison is a distance from the mesh extending outwards and inwards. Not above or below in the Z axis. Hope that makes sense.

[daniel]

I agree with orinn.

And just one remark: for ICP (fine registration), don't set the RMS difference too low (1e-6 is more than enough). Below this value you'll hit the internal computation accuracy limits.