Scanning the object

Post-processing - Overview and Controls

Expand
title Figure A
Image Modified

SAVE YOUR PROJECT WHEN YOU'RE DONE SCANNING!

Once you're done with your scans, you can begin processing them into a 3D model. If you're working with multiple objects and are pressed for time, it's a good idea to leave the processing for later when you're back in the office. Of course, you'll need to keep in mind that if you won't get access to the objects again, you'll need to be extra careful about ensuring you have full coverage of the object. Here is the basic outline of the steps in the process (fig. A):

1. Fine registration (usually done automatically after exiting scanning mode).
2. Editor - base and excess scan data removal
3. Alignment - auto alignment if possible
4. Global registration - possibly followed by deleting frames
5. Outlier removal
6. Fusion - sharp fusion most likely
7. Fix holes & small-object filter - or go back and re-scan portions of your object
8. Texture application & mesh simplification - unless output is just an STL file

Perhaps the steepest learning curve in this whole process is becoming proficient in rotating and manipulating 3D objects. Controls are consistent across tools, so here they are:

• LMB: Click and hold to rotate object around center axis.
  • Double LMB - reset the center axis (signified by red dot)
• RMB: Click and hold, moving up or down, left or right to zoom in or out. (this can also be done with just scrolling, but the mousewheel method is much more imprecise than using RMB).
• Mousewheel: click and hold to move object laterally up and down or left and right.
• Ctrl + LMB: brings up selected tool, click and hold to select region.
• Ctrl + scroll: changes selected tool size
• Ctrl + Shift + Scroll: adjusts plane level (cutoff-plane selection tool)
• Shift + LMB: move only unregistered scan (Alignment step)

Post-processing - Editor tools

Expand
title Figure A
Image Modified

Expand
title Figure B
Image Modified

Expand
title Figure C
Image Modified

Expand
title Figure D
Image Modified

Post-processing - Alignment

Expand
title Figure A
Image Added

Expand
title Figure B
Image Added

Expand
title Figure C
Image Added

Expand
title Figure D
Image Added

Just like in photoshop, you can enable or disable different layers, or in the case of Artec Studio, scans. It's the same little eye icon in your workspace window to the right (fig. A). Before you hit the Align button, make sure that all of the scans that you want to align are enabled. Once they're all enabled, hit the Align button and then click on Auto-alignment (fig. B). If all of your scans auto-align, you're all done! Move on to the next step, global registration. If auto alignment doesn't work, it'll be pretty obvious to the human eye. If the object is too symmetrical and you're unsure if the auto-alignment has worked, try switching to the scan color to see if the surface geometry looks aligned. If you're still unsure, you can either work through manual alignment or complete the rest of the process and compare the finished model with the original object.

In the event that auto alignment just doesn't work, you can do it manually. Recall earlier that I said that you should plan to have at least one scan that has overlap with at least two other scans. This is the part of the process where that can really help. In figure B you'll notice that our two scans are now on the left. One is bold with a blueish circle next to it, the other is regular font with a green circle next to it. The same color scheme can be see in the picture-in-picture screenshot. The bold/blue scan is called "registered" where as the green scan(s) are unregistered. You'll want to set the scan with the most coverage, or the scan that overlaps with two or more other scans, as your registered scan. To do this, simply right click on the scans to change them from registered to unregistered and vice-versa. In our example we are only aligning two scans, but if we were working with more than two, you could have multiple unregistered scans. Registered scans are always visible, unregistered scans only become visible when they are selected. Although you can select more than one unregistered scan at a time, more often than not you'll only want to work with one unregistered scan at a time.

Once you have your registered scan in place, select one unregistered scan. Hold down the Shift key and use the mouse button as you would normally to rotate/move an object. Notice that the unregistered scan is the only one that moves. If you don't hold down shift, you'll move all of the scans, both unregistered and registered, with your mouse. Position the unregistered scan so that it is next to the registered one, with the overlapping portions of each scan visible to you. Recall that you can change the axis of rotation by double clicking. In figure C, I've got both my scans aligned in such a way as to show the overlapping area. Sometimes it can be difficult to tell, so it can be helpful to toggle between the scan color and the texture to highlight both the color of the object as well as its surface geometry.  In this case, there's a dark patch that is a good indicator, and the surface geometry reveals carvings that appear to be the same as well. The more practice you get with manual alignment the better you'll get at planning your scans ahead of time.

Now that the scans are oriented in this way, we can begin to tell Artec Studio where to align the scans. To do this, simply alternate clicking on one scan and then the other, with two successive clicks indicating the two spots on the two scans that should be aligned. You can toggle between the texture and the scan color, as I mentioned, to help you find matching points, as in figure D. Once you've got your points (minimum three sets, in my experience), click on the Align Markers button. Rotate the scans to make sure it looks good, and then either click on the Align button or right click on the unregistered scan you're working with and mark it as registered. Continue on until all your scans are aligned and hit the apply button. You're now ready for the next step: Global Registration.

Post-processing - Global Registration

Expand
title Figure A
Image Added

Expand
title Figure B
Image Added

Save your project.

Under your Tools window, you'll find Global Registration. For more details on what exactly global registration does, consult the Artec Studio manual. Suffice to say that global registration takes each individual scan, no aligned, and turns it into a single point cloud behind the scenes. I say "behind the scenes" because there won't be any one scan added to your workspace. Instead, what you'll see is the Max Error of some or all of your scans change. Ideally, this step makes the max error go down. The ideal for the Eva is around 0.3-0.5, the ideal for the Space Spider is 0.1. Max error refers to the maximum margin of error, in millimeters, on one or more frames. In other words, if you double click on one of your scans, you'll see each individual frame in that comprises the entire scan.

If you're curious about the settings for Global Registration, again I'll refer you to the manual. Default settings have always worked well for me. Once you've run Global Registration, you may see something like figure A: the max errors now say Warning and 0.2 instead of 0.3 and 0.3. Double click on the first set of scans to bring up the individual frames. (To get back to the scans, click on the icon with the little red arrow in fig. B.) Click on the Max error column and toggle the sorting method for the column until you see "Failed" at the top (fig. B). In this case, there was only a single frame that failed during the global registration step. It could be multiple frames. However many it is, highlight them all and delete them. If you're concerned with how man you're deleting, consider the total number of frames in the scan and then decide on whether or not you need to re-scan the object. Once you've gotten rid of failed scans, you can now get rid of frames until you're satisfied with the max error. Once we've deleted the failed frames, resort the max error column until its in descending order. In this example (fig. B), there are three frames that 0.2, the rest are either 0.1 or 0.0. In this case, I would delete the three 0.2 frames so that this particular scan has a maximum error of 0.1: the ideal. This portion is subjective and isn't technically necessary. However, it can yield better results down the line. Use your discretion. If you've only got two scans that only have a couple hundred frames each (unlikely), then maybe deleting 20 frames is too many. On the other hand, if you have six scans each with at least 500 frames, maybe deleting 15 here or 20 there isn't that big of a deal. 

Once you're satisfied, you can run the outlier removal step.

Post-processing - Outlier Removal

Expand
title Figure A
Image Added

The outlier removal step is a simple one, and it can be found under the Tools menu just below Global Registration. This time you'll want to click the little down arrow next to the tool to reveal the tool's settings. Change the resolution to the largest Max error value in your workspace (fig. A). So if you have six scans, five of which have a max error of 0.1 and one of which has a max error of 0.2, your resolution for Outlier Removal should be 0.2. Click apply and let the algorithm run. It's that easy!

Quick tip: if you notice that after the outlier removal step that you've got large holes, you have two options. Your first option is to re-scan the area in question. To do this, undo at least the outlier removal step (fig. A) and just click on the Scan button and start scanning. (Ideally you'll want to undo any frame deletions that you've done as well, but undoing the outlier removal is really the only thing that you really should worry about.) Your second option involves duplicating the scan(s) that cover the area that the outlier removal command deleted. If your scan(s) didn't have enough coverage of an area, the resulting point cloud can be sparse, which can trick the program into thinking that this sparse area is comprised of outliers that need to be removed. To circumvent this, duplicate the scans in question so that the sparse area becomes more densely populated with points. The number of times you need to duplicate scans will vary, so it may take some experimentation. This last tip can also be a way to salvage a scan that you aren't able to re-scan!

Once you're satisfied with the outlier removal results, it's time to create your model!

Post-processing - Fusion

Expand
title Figure A
Image Added

Expand
title Figure B
Image Added

Expand
title Figure C
Image Added

Save your project.

There are three fusion commands: Fast, smooth, and sharp fusion. For the differences between them, consult the Artec manual. For our purposes, we'll only be using Sharp Fusion. It'll be the highest quality model that your scans can produce. Just like with the Outlier Removal command, we'll be getting into the settings for this one, so click the arrow next to Sharp Fusion. Here, we'll be changing to resolution to the same resolution that you used for the Outlier Removal command; in this case (fig. A), it'll be 0.1. (Just like with the alignment step, make sure that all of the scans that you want to use are enabled with the little eye icon.)

Note that sometimes 0.1 resolution for this step will result in a model with too many polygons. A sharp fusion command with 0.1 resolution will also take much longer than the command when it's run at 0.2 resolution. For cultural heritage artifacts, the highest resolution is best. Later steps will allow us to reduce the number of polygons if we so choose, but this is the only part of the process where we can maximize the polygon count. (In this vernacular, higher resolution = more polygons.)

If you're ever wondering if your program is stuck, look at the very bottom. The total memory in use will likely be fluctuating, which indicates that it isn't frozen. Next to that you'll see the command that is currently being run, and next to that is the progress bar (fig. B). Double clicking on the progress bar will open up a new window with just the progress bar displayed, since the bar along the bottom can be easy to miss. This window will also let you cancel a command, although it can sometimes take quite a while to cancel the command, depending on how long it's already been running. Don't try and do anything else in the program while this is running, and if you're working with a particularly large project, it's best to just leave the computer be. Either let it run overnight, or go grab a cup of coffee.

Once it's done, you'll see a (hopefully) gorgeous 3D model with one of the same random metallic colors that the scans are designated with! (Fig. C)

It's time to check to make sure there aren't any holes in the model.

Post-processing - Fix Holes and small object-filter

Expand
title Figure A
Image Added

Expand
title Figure B
Image Added

Before filling holes, run the small-object filter. Under tools, click the settings for small-object filter and select "Leave_biggest_object" and then click apply. This setting is self-explanatory, and the manual can be consulted for the details of the other setting.

There are a couple ways of filling holes. For the purposes of these instructions, we'll go over one of them. Click on the Fix Holes icon to bring up the window. In our example project, we didn't end up with any holes after sharp fusion, so I've deleted portions of the model to show you what these steps might look like. It is almost certainly the case that your holes will not look like perfect circles, and they could also be much, much smaller, so it's good to check with this tool before you're certain.

Figure B shows the Fix holes tool, along with the two holes selected and outlined in yellow on our model in the viewing area. As a side note, you may want to uncheck the box next to "Move camera to selection," as this can be rather annoying. To fill the holes, simply click on "Fill holes" and wait for the command to finish, then click apply if you're happy with the results.

In the case of cultural heritage artifacts, it's very rare that you would want to do this. You would instead want to go back and make sure you've done everything you can to avoid the holes in the first place, either by duplicating scans as described in the Outlier Removal section or by going back and re-scanning the areas that produce holes (also described in the Outlier Removal portion). Exceptions to this are very deep crevices or holes that you know the scanner cannot capture. If you are in doubt, go back and re-scan. The final product should be the highest quality scan possible, so that means either having zero holes to fill, or minimizing the size of the holes as much as possible. If holes are unavoidable, take screenshots and notes so that you are transparent about which objects have holes and where those holes are. You can ask for input from experts as to whether or not the model with the holes is preferable to the model with the holes automatically filled.

Either way, you can proceed to the final texture phase if that's your end goal. Otherwise, export the mesh (with or without holes) to whatever format you need and you're done!

Post-processing - Texture Mapping

Expand
title Figure A
Image Added

Expand
title Figure B
Image Added

Expand
title Figure C
Image Added

Expand
title Figure D
Image Added

Expand
title Figure E
Image Added

Expand
title Figure F
Image Added

Depending on the polygon count of your model, the texture mapping phase can take a very long time. Texture mapping involves assigning a color to each individual polygon in the model, which oftentimes means millions of polygons. You can see the polygon count by double clicking on the fusion model in your workspace (fig. A). Click the back button to go back to your scans. On a computer with 64gb of RAM, I've found that 10-12 million polygons is around the upper limit of polygons that the computer can handle when applying texture. Anything more and the software has a high chance of crashing.

After clicking on the Texture tool, you'll see a window that has each of the scans that you had enabled with the eye icon listed in a window (fig. B). Here you're telling the program which scans to use when applying the texture. It's usually good practice to select them all the first time around. Default settings should be fine. Once you hit apply, don't go anywhere. You should now see Artec Studio saving your project. Wait until it has finished saving your project, because it may prompt you with warning before actually running the texture command. The warning has to do with the polygon count (fig. C). If you're comfortable with the count, proceed despite the warning. Now feel free to leave the computer to run overnight or to grab a cup of coffee, this step can take a while.

If you want to simplify the mesh, there are two options: Mesh Simplification and Fast Mesh Simplification. The former is an automated process that runs based off of certain settings that you input (fig. D, consult the manual for details about these settings). This is recommended if you want to retain as much surface geometry detail as possible. The latter option, Fast Mesh Simplification, allows the user to input a specific polygon (tri_num) count to reduce the model down to (fig. E). This is very handy if you have a maximum limitation, say on Sketchfab or Unity, that you need to be under. It's useful to know that even if you reduce the polygon count from millions to tens of thousands, applying the texture can do an amazing job of disguising the low resolution.

The final product is in Figure F, along with some very basic texture adjustment tools. (Note that for this example I ran mesh simplification prior to the texture tool, which is why it only took 4 seconds to run!)