Fix poisson script. Enhance ball pivoting script

[JimmyDaSilva]

Hi @vrabaud,
I was having quite bad results for mesh reconstruction and decided to have a look at the meshlab scripts.

1. Since the voxelgrid in pcl has been set to 0.005, I think that a ball radius of 0.02 is a bit too much. I suggest switching to 0.01. Also the Laplacing smooth was smoothing too much.
   Here is the results I get. On the left new script, right old script:
   

2. I discovered there is a second script in the code to allow the use of the Poisson reconstruction but it was not complete. The Poisson reconstruction requires that the normals are computed before or you will get silly results:
   
   Once the normals are computed and smoothed you get this instead:
   
   What about we add a new argument to allow the use of this script ?

3. Poisson could do a nice job here if we could close the bottom of the mesh like propose in this issue:
   Closing the loop in object model capture#18. Any thoughts on this ?

4. Once we get linemod to work with texture we could had another script creating the texture and an .obj file. I followed the following video tutorial this afternoon and it works like a charm:
   https://www.youtube.com/watch?v=6wP_e37t7PI
   Gives the following results:
   
   You can also then smooth the colors on the texture file to improve the results.

I would love to get linemod to be user-friendly and use the color gradients for detection. I can work on it but I still need your point on view and experience on all of this.

Thanks again for this amazing library. Looking forward to working on this,
Cheers,
Jimmy

[vrabaud]

1: ok
2: yes
3: I am no expert there, it would require some assumption about your object (not hollow for example). That is a tough one.
4: sure, I need to answer on the LINEMOD thread about using color with LINEMOD but if you find a method that works, we might as well store the colors right now.