What can we help you with?

< All Topics
Print

Photogrammetry

Photogrammetry is the process of building a 3D model from 2D photos. It can be done with any camera, even an ordinary cell phone. This makes it excellent for use in situations where a more expensive scanning solution is out of unaffordable, the object is too remote or aerial scanning, where it’s impractical to use anything else.

1) Take lots of photos.

Photogrammetry starts with the images. The higher resolution, the better. It is important to make sure that the object being scanned has a lot of detail, or has lots of color changes and is not shiny. A flat white board will be very difficult to scan because there are not many pixels that are distinct from one another. A shiny object will change reflections from image to image, meaning that pixel groups will not be consistent between different images no matter what. A tie-die shirt would be perfect; lots of distinct pixels and it is not shiny at all.

2) Feed the Photos to Photogrammetry Software

There are many software packages for this, including Meshroom, Metashape and 3DF Zephyr. Regardless of the software used, they all start by picking groups of unique pixels in the images. These groups are compared against each other from different camera angles, building a map of where they exist in space. These can be referred to as tie points.

3) Build a 3D Point Cloud

The photogrammetry software will then use these tie points to position the pictures in 3D space relative to one another. Then it will use those images to generate a dense cloud of points to build the model from.

4) Build the Mesh

From this point cloud a polygonal mesh is generated. This is what can be displayed in video games or shared easily on the web.

5) Generate and Apply the Texture

In order to give the model its color a 2D image is created from the photos. This is called the texture. Once generated, it is applied to the model to create the final result.

A Note on Accuracy

Photogrammetry can create extremely precise and accurate scans but they will not be the correct size when brought into the virtual world. This is because the process has no way to tell the size of an object, only the shape of it.

For this reason if accurate sizing is important then an object of known size should be included in the scan. This will allow the scan to be rescaled to the correct size in the virtual world.

Previous Laser Scanning
Table of Contents