< Back

An explanation of different 3D scanners

There are a few different 3D scanners which operate with varying levels of complexity. There are scanners that are cheap and portable but produce low-quality results. There are incredibly high-definition scanners that are expensive and only capture objects of a certain size. There are scanners that are great for inanimate objects but terrible for people and animals. I will list the different scanners that are used to capture people but first I will explain what sets the scanners apart.
There are 3 important factors to consider with 3D scanners:

1 - Topology

This is the surface of the 3D model that results from a scan. This is often referred to as the mesh. It is made up of polygons put together in 3D space. You want the resolution of this to be as high as possible with the initial scan. This is particularly important for the face. The topology of the face is what really defines our look. The rest of the body usually consists of parts that are relatively smooth and can be reconstructed even if the topology is damaged or missing. However, its extremely difficult to reconstruct the topology of the face without altering a person's defining look altogether.

2 - Texture

This is the colour that's wrapped over the 3D mesh. It's literally a square image when flattened out. The quality of the texture is what really gives a 3D scan the "high definition" look. Also, the higher the resolution of the texture, the more you can zoom in to your scan while retaining a high level of clarity. Your texture is considered high definition if you can zoom into a scanned human and clearly see stubble, blood vessels in the eyes and even pores of the skin. However, this isn't always necessary for gaming and definitely not necessary for 3D printing which significantly reduces the resolution of the texture when printed on sandstone.

3 - Time

This is the time it takes to scan a whole person. The longer it takes, the more motion distortion you will suffer. The requirements here should be the same as the requirements for taking a portrait photo with a normal camera. The recommended shutter speed for a portrait photo is 1/160. That means the shutter is open and capturing everything for 1/160th of a second. The subject needs to keep still for 1/160th of a second. If you've seen a photo captured by a DSLR camera with a quick shutter speed you'll notice the clarity of the image. Once you decrease the shutter speed you start introducing blurring into your image.

3D scanner types

There are a few common types of 3D scanners out there. I will briefly explain each one using human scanning as an example. Humans are difficult to scan because of their size and because they are animated.

Handheld scanners

These are the kind that Artec and Kinect make. You walk around a person and scan them in carefully. The topology comes out quite detailed, but the texture is very low quality, resembling that of a low resolution compact camera. The major drawback here is the time it takes to scan a person, which is over a minute.

Multi-IR scanning rigs

This uses the same technology as handheld scanners but its an automated process. A person stands in the middle of a rig which is surrounded by around 8 IR scanners. You stand still for at least 15 seconds while these 8 scanners rotate around you. The most number of polygons you can get from these are about 500,000. The difference between this and using one handheld scanner is just the scanning duration. A lot of scanning booths are made like this. You might even see them in shopping centres, being used as a tool to measure a customer's dimensions for easy fittings. These scanning systems also suffer from low resolution textures and motion distortion. You would never take a portrait shot with a shutter speed of 15 seconds!

Single camera photogrammetry

Another process used for 3D scanning is photogrammetry. This is where photos are aligned in 3D space using features in the photos. From there you can produce a mesh and texture. Its basically panorama for 3D. If your nose is in one photo and also appears in the 2nd photo, you can use that feature to match the 2 photos together. The downside is that this is reliant on features, but, the clearer the photos, the more features you can extract from it. With any kind of camera you can take 80 photos of a person from different angles then create a 3D scan just from that. However, a better camera will bring out more features so you will have more accurate topology and a clearer texture. The downside of using one camera is that the subject needs to keep still for a few minutes. This will introduce a lot of motion distortion.

Multiple camera photogrammetry

This is the grand-daddy of them all. This is a setup that relies on around 80 cameras which are all synced together. The purpose of this is to capture the person from all angles within the smallest possible time-frame. Its possible to go down to 1/200th of a second synchronisation between all cameras. This eliminates motion distortion and is more comfortable for the model since its just like a normal portrait shot. You could even capture them while they are in motion. A cheaper set-up might use compact cameras. These will produce a less accurate topology and lower resolution texture. A top-of-the-line setup would use DSLR cameras. The initial topology here would consist of millions of polygons (from 3 million to 7 million) and the texture would have the same clarity as a professional photo taken by a DSLR camera.

Compact cameras vs DSLR cameras

Here is an image of a 3D scan we took using our old rig consisting of 80 Canon A2500 cameras

Here is an image of a scan we took using our current rig consisting of 82 Nikon DSLR cameras