So far, you have used the camera of the mobile phone exclusively for rendering the view of the real world as the background for your models. Computer vision-based AR goes a step further and processes each image frame to look for familiar patterns (or image features) in the camera image.
In a typical computer vision-based AR application, planar objects such as frame markers or natural feature tracking targets are used to position the camera in a local coordinate system (see Chapter 3, Superimposing the World, Figure showing the three most common coordinate systems). This is in contrast to the global coordinate system (the earth) used in sensor-based AR but allows for more precise and stable overlay of virtual content in this local coordinate frame. Similar to before, obtaining the tracking information allows the updating of information about the virtual camera in our 3D graphics rendering engine and automatically provides us with...