Monocular cameras

Search IconIcon to open search

Source:  Cometlabs

  • + Simpler hardware implementation
  • + Smaller and cheapter systems
  • - need complexer algos and software because of lack of direct depth information from a 2D image

How is the shape of the map generated?

  • Integrating measurements in the chain of frames over time
    • Use triangulation method
    • As well as camera motion, if camera isn’t stationary
  • Depths of points are not oberved directly (s. monocular depth perception )
    • estimated point and camera positions are related to the real positions by a common unknown scale factor (scale availability issue)
    • requires camera movement when depth is to be calculated

Adressing the scale availability issue

  • Stereo cameras
  • Introduce a real metric scale (via external scale reference: pre-specified object or sth w/ known size that can be recognised during mapping)