Graphics Reference
In-Depth Information
(a)
(b)
Figure 8.3.
(a) An example LiDAR scan of a large building. This scan contains approximately
600,000 points and the building face is approximately 60m across. Note the characteristic “shad-
ows” on the building due to occlusions by the tree and truck in the foreground, and missing
returns in several of the windows. (b) A detail of the scan.
glass
LiDAR
(a)
(b)
(c)
Figure 8.4.
Characteristic problems with LiDAR scanning. (a) Missing data (thick gray lines)
due to occlusion (back side of circular object) and “shadows” (hidden area of rear wall). (b) False
or missing returns from rays directed at glass surfaces. (c) Grazing-angle errors from surfaces
nearly parallel to the incident ray.
also pass directly through it, returning the distance to an object on the other side.
Also, dark or highly absorptive surfaces may be difficult to scan since an insufficient
amount of light is reflected back to the scanner. Finally, surfaces nearly parallel to
the laser beam often produce poor-quality distance measurements, since most of
the light is reflected away from the scanner; these are called “grazing-angle” errors.
Therefore, the best-case scenario for LiDAR scanning is a scene containing bright,
matte surfaces that are all roughly perpendicular to the laser beam. Man-made struc-
tures like buildings typically scan well (except for the windows, as can be seen in
Figure
8.3
b). Accurately scanning a difficult object like a shiny car typically requires
it to be treated with an even coat of matte white spray-on powder beforehand.
5
There aremany smaller issues as well. For example, the spread of the laser spot over a long distance
can affect the accuracy of the return, as well as the interaction between the color of the laser and
the color of the scene surface. That is, a green laser may be very accurate for a faraway white surface
but give an inaccurate or missing return for a red surface at the same distance.
