(Last update October 25, 1997)
3-DIMENSIONAL DATA EXAMPLES
Some of my students and I have built our own scanning systems over the years and found
that they give us insight into the problems associated with 3-D scanning. We decided to
share some of the original data from some of these prototype systems with the public and
invite you to view them. In doing so we note that the data may contain errors associated
with the system, and in fact, some of the data was obtained during the debug and
calibration of the scanners. There is also some terminology associated with the scan data
which includes:
Structured Light Illumination:
Spatial structured and Time modulated
Not completed.
Perspective distortion
As the distance between and object and the camera lens changes, so does the magnification.
If this variation in magnification is not compensated for then the resulting 3-D
reconstruction will have distortion. However if the variation in magnification is
compensated for then the distortion is reduced and a wider variety of lens systems can be
used. Even with perspective compensation there will be a form of discontinuity between
adjacent scan frames. For object regions close to the camera, there may be a gap between
adjacent frames whereas for object regions beyond the focal plane of the camera, the
adjacent frames will overlap. The less desirable problem is the gaps since information is
lost and can not be recovered.
Intensity Mapping
During structured light illumination, it is possible to store the intensity values along the
light stripe. This data can later be mapped to the vertices. The resulting 3-D
reconstruction has the realism of the original object.
Color Mapping
Not completed.
Range Images
Range images display less detail than intensity maps because the pixel values only carry
range information. That is, a pixel value of black (0) usually indicates the farthest range
while a pixel value of white (255) indicates the closest possible range.
Vertex Only Plots versus Polygon Plots
For vary dense sampling, plotting the vertices (shaded for range or intensity) may be
sufficient for recognizing the object. For a more recognizable image, the empty space
between the individual samples is filled in where the shading of the polygon represents an
interpolation of its vertices (i.e., sample points) shading or color.
Z-buffer for Hidden surfaces
A Z-buffer is commonly used in plotting hidden surfaces. The Z-buffer contains range data
as a function of lateral X-Y position. Before an X-Y point can be plotted, its range value is
compared to that already stored in the Z-buffer. If the new point is behind the nearest
point (Znew > Zold) then the point is not plotted. However if the new point is closer than
the nearest one, the point is plotted and the Z-buffer location is updated with the new
range value Znew.