An experiment was conducted to examine the validity of freely
available photo-based 3D scanning software for generating accurate
3D geometry of a vehicle. Currently, 3D vehicle geometry is
routinely captured using total station survey equipment, coordinate
measuring machines (CMM), laser scanning, or traditional
point-based photogrammetry. While these methods produce
sufficiently accurate results for collision reconstruction, they
have limitations that can affect practical implementation. For
example, manual surveying with a total station, CMM or traditional
photogrammetry are all limited to the production of coordinate data
at discrete, pre-defined points. In addition, total stations, CMMs
and laser scanning devices are also expensive and require a
significant amount of time in the field with the vehicle. In
contrast, photo-based 3D scanning allows a 3D mesh to be created of
a vehicle simply from a series of photographs using a
consumer-grade digital camera.
In this experiment, a vehicle was marked with point targets and
photographed. The coordinates of the targets were then measured
using: 1) A total station; 2) Traditional photogrammetry software;
and 3) Photo-based 3D scanning software. The coordinates of the
targets on the vehicle model produced through the photo-based
scanning process were compared with the target positions measured
via total station and traditional photogrammetry. The mean
deviation between corresponding points on the photo-based scanning
model and the traditional photogrammetric model was 3.2 ± 1.8 mm.
The mean coordinate deviation between the photo-based scanning
model and the total station data ranged between 3.4 ± 1.4 mm and
6.3 ± 3.1 mm.