Neuroinformatics Group

Universität BielefeldTechnische FakultätNI

Active Stereo Vision System

Active Vision considers vision as a process of active data aquisition, i.e. actively pointing the cameras towards interesting regions within the scene, eventually changing zoom as well. 

Therefore, a central part of a typical Active Vision-System is a camera-head which can control some (or all) of the following parameters:

  • Spatial-position
  • Orientation of the viewing-direction
    • Neck-movements (Pan- and Tilt-motors)
    • Vergence-movements of the eyes (Pan-motors for the cameras)
  • Focus
  • Zoom
  • Iris
  • Further camera parameters

In research they often use stereo-color-systems. In context of active vision the system-parameters described above, should be adapted to the actual scene. Under certain circumstances a complete generation of a world-model of the environment can be avoided, in that only interesting parts of the image are analysed. This results in a reduction of information, and so you are in a situation where a realtime-image-analysis is possible with real images from complex environments. Furthermore, a change of the image-parameters could lead to a simply interpretation of a scene, which is done much more robust and faster than with a static system. Possible applications of active vision-systems are in the areas of service robots, quality-management, aids for handicapped and many other fields.