Next: Range Acquisition Hardware
Up: Capturing Dense Environmental Range
Previous: Image-Based Rendering
Figure 1 shows a typical dataset of reflectance and range
data, resampled onto a grid). The projection is spherical (each
column is a longitude where the samples are spaced at equal angles),
as it most naturally matches the rangefinder's rotating mirror and
panning unit.
Figure 1:
The spherical image of reflected laser
light and range data. Each column starts at
60 degrees below the
horizon going up, over the North Pole (middle of the image) continuing
down the other side, until
60 degrees below the horizon is again
reached for a total of
300 degrees per column. The horizontal
field-of-view is
180 degrees, providing a full panorama of the
environment that is only missing a
60 degrees cone towards the floor.
 |
Figure 2:
A reprojection looking toward the windows using
the data from figure 1. The scenery outside the windows
is too far away to reflect enough light and is shown as black.
 |
Figure 3:
A view from above a computer table. A chair,
keyboard and monitor are on the left. Next to the table, on the right,
is a computer with all of its wires shown behind it.
 |
Figure 4:
A view showing the ceiling tile pushed upward,
allowing us to calibrate the azimuth.
 |
The data is shown in subsequent figures (2, 4, and
3) visualized as point-clouds, using the reflected intensity
of the laser light to color each point. The images are captured from
a visualization tool as we move through the data. We use this tool to
explore the data, modifying it with a variety of techniques (described
later in the paper) to eliminate errors in the data. Figures
2 and 3 are rendered from a single set of range data
collected in figure 1. Figure 4 is from an
earlier set acquired in the same room during a calibration experiment
(described fully in section VI-A).
Next: Range Acquisition Hardware
Up: Capturing Dense Environmental Range
Previous: Image-Based Rendering
Lars S. Nyland
1999-02-19