Here we put together some information on the subject of "photogrammetry".

This information is intended to help teachers and students carry out the project days "digital preservation of monuments" help

Photogrammetry... Huh?

Photogrammetry is the fastest and most accessible method to create three-dimensional digital twins.

In short, hundreds or thousands of photos are taken of an object or a scene, which are then converted into a 3D object using software.

The result is then, for example, like this:

So: what do we need?

Requirements and required equipment

The good news first: the costs are minimal, the topic is easily accessible and quickly creates a fascination for students of all ages from the 9th grade onwards.

What we need:

• Cameras - the smartphone cameras of the students are completely sufficient, any existing DSLR or system cameras are all the better
• at least one PC with an NVIDIA graphics card
• The software RealityCapture
• an account at Sketchfab.com
• Know-how how to photograph scenes and objects
• Know-how in dealing with the software RealityCapture

The cameras

The cameras are of course used by the students to take photos of an object or a scene.

The higher the camera resolution, the more details the later 3D object can reflect.

Modern smartphones with their 12 megapixel sensors or better are completely sufficient for our purposes.

PC with NVIDIA graphics card

This may be a hurdle for some schools: we need a "gaming PC". We need the Nvidia graphics card because the "RealityCapture" software we use requires it.

The good news: we don't need the latest model. Even old cards like an NVIDIA 760 GTX will do just fine.

RealityCapture

There are various software for the topic of photogrammetry - we rely on the top dogs RealityCapture.

From our experience, this software not only delivers the fastest and best results - the license model also suits our project very well.

We can use the PPI (pay-per-input) license model. You only pay when you export a 3D model (e.g. export to Sketchfab) and you only pay for the photos that were actually used to create the geometry of the model.

On the software's website there is a calculator for expected license cost

If we assume the case that we took 300 photos of 12 megapixels each for an object, this model will cost us 2.25 $

Payment can be made conveniently by paypal or credit card.

Sketchfab

Sketchfab is almost the InstaGram for 3D content.

3D models can be uploaded to Sketchfab, shared, commented on and even sold.

One should see Sketchfab as a separate social channel that can generate further reach.

So the school certainly doesn't make a mistake by creating an account here, collecting the generated 3D models here and then integrating this content into their own social media activities.

How should the photograph be taken?

Photogrammetry requires a slightly different view of the scene than classic - artistic - photography.

Basically you look at the object in question and analyze the geometric complexity.

The following applies: Areas with great geometric complexity must be photographed more frequently and from more perspectives than less complex areas.

It also makes sense to take photos from varying distances.

Every point of the object should be visible in at least 3 photos from different perspectives.

It makes sense to first capture a whole scene or object with a few photos, and then gradually map the identified details with more photos.

This makes it easier for the software to later assign detailed photos.

The software must be able to locate each photo in the scene.

This means that a fairly high overlap between the individual images must be achieved.

As a rule of thumb: there should be at least 70% overlap between images and for orbits around an object there should be no more than 10 degrees between 2 photos.

As I said, these are rules of thumb – each object needs individual consideration.

weather and light

Both shadows and glaring light are bad. The best day for outdoor photogrammetry is a windless, overcast day with soft, diffuse light that doesn't cast shadows.

wind

Calm wind is important because the geometry of the scene to be captured must be stable during photography:

Bending branches or bushes in the wind, leaves being blown around, or the like will change the scene and give bad results or make results impossible.

What is suitable?

Photogrammetry is now a very stable method. However, since it is a light-based process, there are limitations:

Objects - or parts of objects - that are reflective or transparent are not suitable.

One can try, but such areas are usually mapped unsatisfactorily.

(Of course you can edit this afterwards in a 3D software)

RealityCapture

RealityCapture (or RC for short) is our tool of choice when it comes to photogrammetry.

The following isn't an RC tutorial - we're just going through the essential process steps quickly here.

If you want to dive deeper, take a look at the excellent Youtube channel. There is also a Facebook Group with a very active and supportive community

At the beginning you should get an account on the homepage of RealityCapture do. Here you can buy "credits" to pay for the export of the data.

From the user's point of view, the process with RC is roughly divided into these steps

• Adding the photos
• alignment
• Generation of the 3D model
• Texturing the model
• Export of the model – e.g. to Sketchfab

Adding the photos

unsurprisingly, the project begins with the addition of the photos.

This can be done in the classic way via an "open" dialog or by dragging and dropping into the left-hand area of the screen.

alignment

(Alignment->align images)

The next step is alignment. Similarities are sought in the images and the respective positions are determined from which the individual photos were taken. In addition, a point cloud is created, which already gives an idea of the 3D object.

Depending on your computer and the number of photos, this process will take a few minutes.

On the left side you can see 115/115 – that means: RC was able to locate all 115 of the 115 photos in space. On the right side you can see the object's point cloud and the camera positions hovering above it. (In contrast to the example above, the camera positions here are quite disordered - but still good enough to cover the entire model)

Generation of the 3D model

(Mesh model -> normal detail)

Next, the 3D model is generated. Ie we get a geometry that is made up of many small triangles.

This step takes significantly longer than the alignment. For large models (thousands of high-resolution photos), this process can take hours to days.

In our example with only 115 photos, it will again be a few minutes.

now you can already see the finished structure. The geometry is constructed.

Texturing the model

(mesh model -> texture)

Now the model has to get its colors.

Export of the model – e.g. to Sketchfab

The model is ready - now it should of course also be shown.

RC offers various options and formats to export the finished 3D model.

A very convenient option is to upload the model directly from RC to Sketchfab.

To do this, your own Sketchfab account must be connected to RC once.

In addition, this step now incurs costs: RC checks which images have been used and calculates the license price on this basis.

If you bought credits as described above, the costs are simply deducted from your credits and the model is then uploaded to Sketchfab.

(For this example, the costs are 115 photos of 12 megapixels each: 354 credits or 0.86$)

And this is what it looks like on Sketchfab:

Conclusion

With a simple camera, an old gaming PC, RealityCapture and Sketchfab, a school class can digitize cultural heritage in a short time and at low cost.

These can be very different objects - if you are looking for inspiration, you can take a look at our Sketchfab account:

https://sketchfab.com/3dtwins/collections

As a further step - and perhaps in upper school classes - it is conceivable to use these 3D models in Blender or the Unreal Engine. In this way, renderings, films or VR experiences can be created with real existing objects.