Parallel & Planar Tracking

Parallel

Currently, what we’re getting in mobile augmented reality apps like Layar, Wikitude, Junaio and the other usual suspects is a run around. Essentially, these apps are blind. They can’t see anything that you see through your phone’s camera. Instead, they get your locality from the GPS satellite system, which is up to 9m wrong, then the direction you’re facing according to the digital compass, gyros and accelerometers in your handset, which are not as accurate as they could be, and cross references all of that with an information database on the Internet that tells it what should be in front of you. Small wonder why it doesn’t work very well.
    There is, however, an area of research called computer vision which is having a very good stab at getting over this. Georg Klein is one of the top minds in this field when it comes to augmented reality. There are three kinds of tracking for augmented reality.”
Tracking is the name given to an AR application’s attempts to recognise and follow the physical objects of a scene. “There’s marker-based tracking which is what the likes of the ARToolKit looks for through you computer webcam. It asks for that specific shape to be recognised. It’s limited but idiot proof. “Next is known texture or pattern recognition where the software might look for a magazine page, CD cover, Lego box, something known. It’s trained to recognise that thing very quickly. It’s used commercially but it must work on geometry that’s known in advance. “Finally, there’s what I do but it’s quite tricky. In fact, it’s so tricky that it’s not yet at the level where anyone can really use it. I can use it. I can show people a few things and get them to use it, but it’s no good if I put the technology out there and hope for the best.” What Georg has come up with is a camera tracking system for augmented reality which is able to model the physical shapes, lines and depth of an environment on the fly. It’s something he calls PTAM - Parallel Tracking and Mapping - and it requires no markers, no pre-maps, known templates and not even any of your mobile device’s inertial sensors either. It’s based on SLAM which is a system of Simultaneous Localisation and Mapping that’s huge in robotics. It’s so people’s little creations can go anywhere without getting stuck in a corner. We adapted the SLAM algorithms to work for AR.”
           PTAM is freely available to download (http://www.robots.ox.ac.uk/~gk/PTAM/) and use but, as Klein himself says, it’s not ready yet. Tracking is fundamentally and mathematically difficult. The world is three dimensional but AR only offers a flat projection of that to work with. Interestingly enough, this is why SLAM appears to work better with stereoscopic camera set ups or something like the Microsoft Kinect which has 3D depth sensor technology as well, although Klein assures us that this isn’t what he’s working on for Microsoft.

Planar

Planar Tracking is used within many areas of our products including:

• Perspective matching for screen inserts
• Linked roto shapes for matte/mask creation
• Image and camera stabilization

Planar, not point tracking.

Traditional motion tracking tools require that you locate “points” that remain consistent throughout the entire shot in order to track movement. This is itself a difficult task, especially when tracking a shot that was not originally designed to be tracked. If you wish to also track rotation, perspective and shear you need even more clear and consistent points to track.

Imagineer’s Planar Tracker doesn’t employ point trackers, but tracks an object’s translation, rotation and scaling data based on the movement of the user defined plane

Vuforia is an Augmented Reality Software Development Kit (SDK) for mobile devices that enables the creation of Augmented Reality applications.^[1] It uses Computer Vision technology to recognize and track planar images (Image Targets) and simple 3D objects, such as boxes, in real-time. This image registration capability enables developers to position and orient virtual objects, such as 3D models and other media, in relation to real world images when these are viewed through the camera of a mobile device.