Fra Robin

Revisjon per 1. sep 2014 kl. 11:04 av Stiansel (Diskusjon | bidrag)
(diff) ← Eldre versjon | Nåværende versjon (diff) | Nyere versjon → (diff)
Gå til: navigasjon, søk



Stian Selbek

Description of master project

3D depth perception is an exciting new development area that the camera industry is working on. It is still uncertain which method will become the market leader. One of the candidates is stereo vision where two image sensors operate in parallel and the distance is measured using geometric calculations.

Stereo vision would be a candidate for sensing solutions in the automotive industry, for e.g. self-driving cars. In this context some important challenges are the dynamic ranges involved and quick movements in the scene (combined with rolling shutter technology this can lead to distortions in the image). Further, there is a need for real time control and thus high speed processing is crucial.

The project goal is to design a stereo camera system that measures the distance to each point (pixel) in the scene. Processing of image data from the two cameras can be done in real time using FPGA or SW. Possible performance focus areas of the design are high speed processing, high speed movements, and high dynamic range. At the start of the project it will be necessary to investigate how much off the shelf technology can be used.

Milestones of the project include:

  • Investigation into previous research on stereo vision. Writing 10-page essay resuming relevant work.
  • Investigate off-the-shelf solutions / components and existing designs
  • Implement a stereo vision system
  • Implement algorithms and propose improvements according to performance focus.
  • Performance analysis


Personlige verktøy