Figure 8.2 - Local registration, remote recognition and tracking

According to the MAR component classification scheme in Clause 7, this system class has the following characteristics:

• 
  Pure sensors – visual
  
• 
  Real world capturer – visual, 2D video
  
• 
  Recognizer – image patch, recognition event, remote
  
• 
  Tracker – image patch , spatial event, remote
  
• 
  Space mapper - spatial
  
• 
  Event mapper - visual
  
• 
  Execution engine – local, 2D + t
  
• 
  Renderer - visual
  
• 
  Visual display – 2D mono
  

The Device sends video frames in a format that can be specified by MAR experience creator (from a local camera capture, a local video track or a remote video resource) to a Processing Server that is analysing the data and detects one or several target resources that stored in its local database. The server returns the position and size of one or several target resources detected in the frame, as well as the augmentation content (virtual objects, application behaviour). By using position and size, the device will crop the target images from the frame and use them for local tracking. The content specified in the Information viewpoint is:

• 
  URLs of the Processing Servers.
  
• 
  URL to the video stream (a local camera, a local video track or a remote video resource).
  
• 
  The format in which the video data is sent to the processing server (raw/compressed image or the corresponding video frame descriptors).
  
• 
  Optional: a 2d region (in the video frame) to be considered in the recognition process and delay constraints.
  

In addition, a communication protocol has to be implemented between the MAR Execution Engine and the Processing Servers.