|
(See ISLA for the new IAS website. Most maintenance activities on this old website have stopped.)
Motion Estimation
|
Motion Estimation and Interpretation with Mobile Vision Systems
|
Objective
In this project we investigate motion estimation and interpretation techniques using a camera
based sensor.
The idea behind this project is the development of a digital rear-view mirror.
This device should inform the driver when an intentional lane-shift is unsafe.
Vehicles driving behind our car limit our motion possibilities.
We observe their motion with a single camera.
The objective is to estimate and interpret this motion, using real image data.
Research group members
ir. M. B. van Leeuwen
prof. dr. ir. F. C. A. Groen
Funding
Informatics Institute
Research Achievements
Within this project we address the following issues:
- egomotion estimation & stabilization
- vehicle detection & tracking
- estimation of the Time-to-Contact & lane-shift detection
Handling the information provided by the camera usually consists of two steps.
First the motion in the 2D image plane is estimated (2nd topic in the list above).
The next step is to interpret this motion and estimate the motion in the dynamic 3D-world
(3rd topic in the list).
An important issue when dealing with mobile systems is that the egomotion of the camera can
seriously affect the results of these two steps (1st topic in the list).
The mean-viewing direction of the camera needs to be known. Furthermore, additional motion
components are induced to the observed motion field by disturbances like e.g. small vibrations
of the vehicle or steering actions of the driver. To enable efficient tracking and accurate
interpretation of the recordings of the camera, these disturbances have to be taken into account.
More information
Home Page of Rien van Leeuwen
|
