Assignment #5: 3D Scanning and Printing

Back to main page

3D Scanning Tips and Tricks

Note: These comments are relevant mostly when you are trying to scan an object that has fine details, and its you want to keep them in reasonable resolution and not lose them. The tricks that John taught us in the lab session (smoothing out the polygons and so on) may be faster, but I've noticed that all the smoothing and noise reduction also removes the finer details.

Scanning the Object

• Light and Shadow

   

  • Try to position the object in a way that will cause as little shadows as possible. Shadows can lead to holes in the scanned object, and lost details. Its best to position the wider parts of the object so that they won't obscure at the bottom, if possible.
     

  • You may want to add additional light, especially if parts of the objects cast a shadow on other parts.
     

  • If you do use the lamp, make sure to position it in such a way that it won't create additional shadows.
     

  • Also, make sure the light shines with equal intensity over the object. If some parts are more bright than others, I think that the scanner adjusts its exposure according to the available light, and this will cause the less-illuminated parts of the objects to come out too dark.
     

  • Play with the preview option before you scan. Repositioning the light by just a few inches can create a big difference in the scanner output.

Editing and merging

• General
   

  • Be especially careful when selecting and deleting point. You are setting your selection boundary on a 2D plane, but it cuts across all of the depth dimension. You may intend to select only the points that you see on the front of the object, but if you are not careful, Geomagic may also select any points that are behind them, at the back of the object.
     

  • Save often, and with alternating file names (at least 2-3 versions in my opinion). Sometimes you discover at a later stage that you have deleted more than you intended, and created undesired HOLES in your object. See the the bullet above. And yes, I'm talking from experience... P.S. Don't forget to delete your alternative file names once you are finished.
     

  • I've found it useful to create some test-merges (Points-->Merge in the main menu) every now and then, while doing the cleanup of the scans. The thing is - you can only fill up the holes AFTER you've merged, but the size and shapes of the holes depend on the image points that you've merged. After you've merged the object and started working on its polygons, you can't go back to playing with the scanned points. What I found useful was to create the these "test-merges", where I would see how smooth my images are, and how the major holes are filled up. Then I'd go back to the scans and correct these areas. These temp-merges can be deleted after you don't need them anymore.
     

  • A very important tool in Geomagic is the option to show/hide layers. In the left side of the screen, you have the list of scanned images and also any merged image that you create. Right clicking on any image will give you the option to hide/show it. This enables you to see exactly what each scan contributes to the merged object.
     

  • Geomagic gives you about one undo step, and that doesn't always work. Work slowly, and after each step you do, see if you got what you intended. If not, undo immediatly, or forever hold your peace.

   

• Clean Up Your Act BEFORE You Merge
   

  • This is what happens if you merge your scans before cleaning up the object surfaces:

    
     

  • This is caused because of overlapping parts of the object, that have been scanned in 2 or more angles. The 2 views are not exactly at the same spot, so there is a short distance between the 2 overlapping parts. So how do we fix it?
     

  • If you just use the noise reductions and smoothing functions, you'd be losing the noise but also the details you do want to keep. The solution I found was also delete some of the duplicate information on the image itself before merging, and not only clean up the obvious "noise" that's around the image   

          .
   

  • Hide the merged object, and bring up the relevant scans. I found it easiest to hide all scans, and then work with only 2 visible ones.
     

  • You'll see that in the areas where the rough surfaces are, the scans have overlapping points. Delete these points from one of the scans.
    I've used the following rules of thumb:
     

    • If one layer is denser than the other in an area - delete the less dense one.
       

    • What will probably happen is that you'll delete the obvious less dense areas from each scan, until you reach a stage where there are areas where they have a similar density. Just decide which scan's points you want to keep.
       

    • If you can, try to make the borderline between the two laters at an area where there is a change in the surface (color change, border between heights, and so on) For example, in my example I tried to make the edges along borders between the face and the hat, or at the side of the face where it will be less noticable. It may not be always possible.
       

    • I usually tried not to delete points at the border edges (perimeter?) of the object, because we don't want to lose information about them.
      
       

  • Here's an example:

                  Merged object:

                 

   

                  Scanned points overlaid to see where the roughness occurs:

                 

                  New merge, after cleanup:

                 

   

  • Another example, before and after:

• Merging and Filling Up Holes
   

  • When you merge, you are given the option to select a spacing value. The default is somewhere around ~0.0036 I think. I've seen John use  0.0009 as a value. I've found it outputs smooth objects, but a lot of the details were lost. Again - if you have a simple object and details don't matter - I guess its better to use the simpler merge. I've played with some values, but eventually remained with the default setting.
     

  • The different values affect how much holes you'll have in the merged image, and they also effect how the hole fill-up will look like.
    Here are some examples:

          Default spacing (~0.00036)

         

                    0.005 spacing (smaller hole)

                    0.008 spacing (even smaller hole)

• • BUT, when I actually started to fill up the holes, the biggest one filled up in the best way, because that merge had kept the most details about the original edges of the hole!!!
     

  • Filling up the holes: There are 2 types of fill "polygonal fill" and "flat fill". Polygonal is supposed to look the best, but I've found that in certain situations the flat fill looks better. Test both, especially if the holes involve complex surfaces.
    Example:
    
    Polygon fill (NOT SO GOOD HERE):

             

          Flat fill (Much better shape. The red color is just because its still marked as "selected"):
   

             

  
   

  • And here is the final result:

Notes:

• If when trying to export to .stl the progam says its not a closed surface, and you think you filled up all of the holes already, go to fill holes again, and this time do an automatic fill. There may be some microscopic holes you missed. Make sure to save before doing the auto-fill!!!

3D Printing

I printed with the ZCORP macnine:

Left: original object
Middle: STL version, original size, single color
Right: vrml2 version, enlarged print, colored

Some more shots:

Notes:

• When "excavating" the printed object out from the print bed, make it as smooth as possible with your hands. The colors should come out better after putting on the wax, plus it will feel smoother to the touch.