Scanning

Destructive

This type of scanner is somewhat atypical because it’s essentially a milling machine with a camera. At the beginning, the measured object needs to be cast into a block so that auxiliary material perfectly fills all cavities. The color of this material must be contrasting compared to the color of the scanned object. Such prepared part is mounted on the mill bed and thin layers of constant thickness are gradually milled off. Each newly revealed layer is always photographed and the image saved for later processing. The result is therefore a set of 2D photos with stored information about what Z height the photo was taken at. Software on each photo at the color transition of the cast object and auxiliary material extracts the boundary curve. This curve is represented as points in the plane. If curves from all milled levels are connected, then we get a 3D point cloud.

Non-destructive

Non-destructive contact scanners include all, unlike destructive methods, the object is not damaged during digitization. Contact 3D scanners examine the object’s surface using physical tangible touch. While the object remains at rest attached to the base, a positioning arm on which a point or ball probe is mounted allows the user to pointwise capture 3D data from the physical object.

Magnetic Scanners

We can divide them into scanners with a magnetic probe or scanners using magnetic resonance. By using the second mentioned type of device, we can obtain information about the internal geometry of components. This is a non-destructive scanner working on the same principle as classical magnetic resonance used in healthcare. Devices are usually mobile and are used, for example, to check pipelines, boilers, or other closed vessels.

Transmissive Scanners

A representative of transmissive scanners are scanners using computed tomography (CT) technology. Just like scanners using magnetic resonance, this type of scanner can obtain data about the internal structure of the examined object. X-rays are used for information transfer. Unlike healthcare versions of CT, this use employs higher radiation intensity. These devices are still relatively rare, which is also proven by the fact that in the Czech Republic there is only one specimen.

Reflective Scanners

This category includes acoustic scanners (e.g., sonar), laser, but primarily optical. Optical scanners are the most widespread and most commonly used branch of 3D scanners. This results in the greatest number of various technological solutions and thus also further division.

Optical – Active 3D Scanners

Active optical methods are further divided according to what physical property of the given radiation is used for calculating the spatial coordinate of a point

Reconstructing Foot Model

Required models are foot_scan.bin and foot_reference.stl. The foot model is from the portal Thingiverse, CC BY-NC 3.0 Voodoo Manufacturing.

In the file foot_scan.bin there’s a point cloud created by scanning the foot model. Open it in the CloudCompare application.

To work with the model, you need to select it in the upper left part of the program (DB Tree). The checkbox is for showing (or hiding) the model. The selected model is highlighted.

For reconstruction we’ll first use the entire cloud. For weaker computers this is not recommended; the cloud has about a million points and it could take a long time. Later we’ll show how to reconstruct a model from a smaller number of points.

First you need to calculate normals using Edit → Normals → Compute. For our purposes, default values will suffice.

Sometimes normals are calculated reversed (they’re displayed in black). In such a case, they need to be inverted. If we didn’t do this, the reconstructed model would be turned inside out. Normals are inverted using Edit → Normals → Invert.

When we have normals, we can use Plugins → Poisson Surface Reconstruction.

If the cloud is too large, we can subsample it before reconstruction: that is, get only part of the points. For our foot about a hundred thousand points suffice.

Sometimes it happens that the resulting mesh is too grainy. It’s possible to "to taste" smooth it using Edit → Mesh → Smooth (Laplacian).

You can also load a finished mesh into the program from the file foot_reference.stl (just open the file).

For clarity of mesh display you can use Plugins → P. C. V. (Ambient Occlusion).

When we position the reconstructed and reference mesh at the same place, we can compare them. In the exercise we’ll show this; if you’re reading materials from home, you’ll learn more in the next example.

Garden Model: Registration of Two Scans

Required models are garden1.bin and garden2.bin, downloaded directly from the project CloudCompare, GPL 2+.

Brief sequence of steps (assumes watching video or attending exercise):

Garden Model: Segmentation, Alignment by Reference Points

Required models are the same as above: garden1.bin and garden2.bin.

First we need two scans that are not so similar as above. To demonstrate this, we’ll cut out only the part with soil using the Segment tool.

The tool is controlled with left mouse button; for confirmation use right button. Then select the cut-out polygon symbol and confirm with checkmark. The operation creates two new clouds in the tree.

Then you need to use the tool Align two clouds by picking (at least 4) equivalents point pairs.

Then you can continue as in the previous example.