Hey, so today i’m going to be learning some more on photogrammetry, and by using my old post where I did some research on it i’m going to learn more about it as what i wrote about last time didn’t go into much detail. One thing I did talk about was how photogrammetry has been used in the battlefield and battlefront games of recent, Dice developers have actually used photography to capture both characters and 3d landscapes and assets for them to use in their landscapes.
To re-iterate on what photogrammetry is, it’s essentially when someone takes a series of images with a camera which is then put into a computer where the computer figures out a 3D model based of off all the images used. Of recent there are many different programs that are able to do this like “Autodesk Remake” Once the 3D model has been created it is created with quite a lot of polys and must be cut down for a 3d program to be able to comprehend it.
There are two ways for photogrammetry to work, one being the normal way of taking 36 images all from a different angle going in a 360 and taking images along a specific route both with camera in hand or with a drone. A series of images must have at least 36 images that form a 360 degree of the object being taken, sometimes it is worth it to do a few more rounds with more images at different heights so that the computer can figure out a better looking model from more angles.
- Take photos in even, soft lighting – cloudy day or in shade works well.
- Take photos in rows where each photo overlaps (more overlap the better)
- Ensure that you have an angle of every part of the object you want to include.
- Doesn’t work well on shiny or transparent objects
To get us started its best to make sure that the lighting settings are right, it has to be a decently dull area for a really good 3d model to be created. There are 3 things to also look out for on the camera those being the field of view, the focusing and the exposure which makes the photogrammetry work really well.otherwise it won’t gather the best results.
Some best practices from dices’ photogrammetry slide, is overlap. What this means is that when taking images they have to overlap so that all angles are covered, this helps with the cleanup and produces a nicer model. Good lighting. Its needed, but it might be a good idea to stop it from giving cast shadow. Avoiding objects with more intricate detail because of the conversion process which might not end up very good.
Some of the workflow from dice devs, once images have been captured include removing vignette and keeping lens distortion. Baking at high gives the most usable and manageable results whereas baking at ultra is best for terrain because of how good of a model it will produce.
The evolution of photogrammetry is pretty interesting (but i also didn’t go through)