Self Localization
Self localization takes place using the omnidirectional camera. Raw camera data at 1600×1200 is converted into 1152×240 panoramic images, taking into account the curvature of the mirrors. Once completed, vertical edges within the field of view are identified based by searching for rapid changes of color.
Once edge detection has been completed, the distance between edges can be used as a measure of angle between those edges, which then gives the robot's position within the environment through triangulation.
Media
The following files demonstrate Stevi's localization module. In order to play these files, you'll need to have Apple QuickTime 7 or later installed on your system.
 |
Edge Detection |
| This video demonstrates the edge detection algorithm used for localization. The top half of the video is the resized input from the omnidirectional camera, while the bottom half of the image indicates detected edges by a vertical bar. | |
 |
Localization |
| This video demonstrates localization within a predefined environment using the omnidirectional camera and edge detection. |