Tutorial: Merging multiple scans

5.3. Tutorial: Merging multiple scans

Problem

In some cases more than one or two images from underneath the object are needed or are desirable to fully capture the shape or texture of the object.

The natural way to handle such an object is to take two sets of photographs, turning the object over in between. In normal operation, 3DSOM Pro assumes that the object does not move relative to the mat, but this is not the case, so the second set of images has to be aligned with the first set.

Solution

The Align Wizard is used to help the process of merging the two sets of images together to form a single project. 3DSOM Pro needs to determine how the object has moved between taking the two sets of photographs.

This can be automated by building a partial model from one set of photos and letting the software compute how the second set is oriented relative to the first set. Any number of additional scans can be merged in this manner.

Starting Point

Create a new project and load in both sets of photographs and generate masks for all the images. 3DSOM Pro is going to use a model created from one set of photographs to align the other set of images against, so choose one set of images, usually the larger set, reject all the other images, and build the model from these images as normal.

TIP:	If the object has few geometrical features it may also be worth generating a rough surface texture at this stage to help with the alignment.

Step 1

Select the second set of images (it doesn't matter if they are currently marked as rejected) and choose Images > Align Images... menu item. This will open the Align Wizard. All the other images now appear locked (greyed out) in the Thumbnail Window.

Select the "Merge Group" option from the wizard welcome page — all the selected images are taken with the object in a fixed position against the mat and can therefore be aligned in a single step.

Choose Next > button to move onto the next page.

Step 2

At this point, look at View Model Window — it displays the current model semi-transparently over one of the selected set of images. You can rotate and move the model as normal. Use the Next view button (and the Last view button) on the View Model toolbar if you want to move back to where you were previously.

You can use either the "playback" buttons (First image button, Previous image button, Next image button and Last image button) in the wizard to move to a different image. When you change to a different image, the model will be rotated so that it remains in a consistent orientation relative to the second set of images.

TIP:	You can also double-click on an image in the Thumbnail Window or drag an image from the Thumbnail Window onto the View Model Window to switch to this image.

Figure 5.3.1. The same orientation of the model seen from two different cameras against their respective backdrop images.

Step 3

As long as there are enough images in the second set and they have been properly masked, 3DSOM Pro will be able to align the object automatically. This is a great advantage, since it is tedious to get the alignment right by hand - it can look good from one image, but from a different image can be far out of alignment. Use the Guess button to get 3DSOM Pro to pick an initial alignment.

TIP:	Switch between View wireframe model (F5), and View flat shaded model (F6) or View textured model (F7) display modes to help show the model up against the photograph.

Sometimes, particularly with fairly symmetric objects, 3DSOM Pro will be unable to tell which way round the object is — each time you click the Guess button it will cycle through a series of possible alignments.

Figure 5.3.2. Two different guesses for the initial alignment — the second is clearly the closest.

Don't worry if the match isn't perfect at this stage - choose "Next" to proceed to the optimisation stage once you are happy the orientation is roughly correct relative to a couple of images.

TIP:	If none of the guesses came close, manually rotate and move the object until it is roughly aligned.

Step 4

During the optimisation stage, 3DSOM Pro attempts to make minor changes to the orientation and position of the object whilst checking that it matches the silhouettes of the chosen images. You will see the object "vibrate" during this process and you can check the alignment against different images as before.

You can stop when you are happy with the results. Use Last view button from the View Model toolbar to revert to the previous alignment.

Geometry Accuracy

You can improve the optimisation by choosing one of three options that best describes the current geometry and the new images. Since the software attempts to match the silhouettes in the new images to the existing geometry, it needs to know whether the existing geometry is accurate or might have major discrepancies. If you are happy with the geometry and have just taken the extra images to improve the texture, then select "old geometry better". If the extra images have been taken to improve the geometry, then you will probably be best off choosing "new silhouettes better".

TIP:	Check the alignment using images from cameras that were widely apart from each other since this is more likely to show any problems quickly.

Step 5

Closing the wizard will fix the alignment of the second set of images relative to the first. They can now be used in the project in the same way that the original set of images were. If you took the extra set of images to capture more shape information, you should continue to rebuild the geometry then the textures. If you are already happy with the geometry you can skip on to just regenerate the textures.

< Previous topic
(Creating or editing masks in an external application)

Next topic >
(Taking the right photos)