Pages created 1995-2000
 |
OFS - Optic Flow Sensor ... or what´s got an MPEG processor to do with Robot Vision? |
The term Optic Flow refers to the motion in video images which is represented by a two-dimensional vector field. Here at the lab we have developed a PC board that can estimate this vector field from a video stream in real-time (up to 525 vectors in 40ms). Because of the computational costs of the underlying correlation operation, a specialized MPEG-Processor (MEP, Motion Estimation Processor) is used. An additional processing unit (realized by FPGA technology) augments each vector with a confidence value. Current research topics include the adaption of common optic flow techniques to the pecularities of this approach with special regards to autonomous vehicle navigation.
Current applications are:
- Detection of moving objects
- Detection of static obstacles
- Determination of the egomotion and local path planning
- Definition of ROIs for the initialization of snakes
- Matching of stereo images
- Matching of current images with reference pattern
