! Ceiling Mosaics
MINERVA: Carnegie Mellon's Robotic Tourguide Project Minerva's Image

Ceiling Mosaics

When the museum is busy, people tend to crowd around Minerva on all sides, blocking her laser sensors which are only 50cm above the floor. This means that Minerva has a great deal of difficulty in being able to tell where she is, or localizing herself. An alternative method of position tracking is needed -- the different patterns and structures of the ceiling of the museum are used for this.

Minerva has a camera mounted near the top of her base, pointing directly upwards, imaging the ceiling. Two example images are shown below.

A single image of the museum ceiling Another image of the museum ceiling

The above images can be enlarged by clicking on them.

Based on the current image of the ceiling, Minerva can identify where she is. For this to work, a pre-existing map of the ceiling must be computed. However, no camera can take an image of the ceiling in one shot. A mosaic of single images is therefore constructed.

To construct the map, Minerva is driven around the museum while it is empty, continuously acquiring images, and labelling them with position information from Minerva's localization routines. A stream of example images is shown below.

A series of ceiling images

The images are then stitched together to form an image of the entire museum ceiling, which can then be used for localization.

The ceiling image

An 800k movie of the localization process using the ceiling mosaic is available.

Home The Java-Based Image Stream About this project