ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002
ON THE CALIBRATION OF MAPVISION 4D SYSTEM
Ilkka Niini
D.Sc. (Tech.)
Oy Mapvision Ltd, Tietájántie 10, 02130 Espoo, Finland - ilkka.niini@mapvision.fi
Commission III, Working Group III/1
KEY WORDS: Adjustment, Bundle, Calibration, Close Range, Measurement, Photogrammetry, Quality, Real-Time
ABSTRACT:
Mapvision 4D measurement machine uses video cameras and a light projector to quickly make a precise 3-D point cloud from a
given small-scale 3-D object. The use of more than just two cameras makes it possible to control the quality of the measured data,
which makes Mapvision 4D more flexible than conventional 3-D measurement machines. The calibration of this kind of
photogrammetric machine requires that all possible effects due to the optical, electronic, and mathematical transformations from the
3-D space to the video images have to be taken into account. The calibration is made using a free-network bundle adjustment, which
is constrained using known distances and points on precise planes. With the first prototype, an operational measurement accuracy of
t5 um is achieved.
1. INTRODUCTION
Mapvision 4D is a new optical 3-D coordinate measurement
machine for on-line reverse engineering and quality control
applications, e.g., in industrial production of small objects. It
uses photogrammetry to measure a 3-D point from the images
of the point. In the first prototype of the machine, the effective
measurement volume is about 200 mm x 200 mm x 100 mm,
and the object accuracy is about +5 jum. The measurement
speed is currently about 80 points per second. The measurement
system contains a standard PC with one or more digitising
cards, four or more CCD video cameras (current resolution
768x576), a light scanner capable to project multiple light
points on the measured object surface, and a rotating table
allowing full registration around the object. The use of more
than just two cameras makes it possible to control the quality of
the measured data. The object is measured in a dark cabinet.
The controlling computer and necessary power supplies are in a
separate enclosure (Figure 1).
Figure 1. Mapvision 4D.
2. MEASUREMENT PRINCIPLE
The measurement using Mapvision 4D is based on principles of
photogrammetry. First, the system is carefully self-calibrated,
by determining the perspective transformations from the object
space to the digitised video images using collinearity condition,
simultaneously adjusting the unknown interior orientations and
additional parameters for lens distortions. Once calibrated, it is
possible to quickly compute the 3-D coordinates of any object
point that can be seen in the cameras by using intersection. The
final result is a dense 3-D point cloud representing the measured
object surface. The calibration needs not to be repeated until it
is accidentally broken or needs to be updated. The effect of
thermal expansion and mechanical strain are also minimized
using carbon fiber composite material in the mechanical
construction of the machine.
In the on-line measurement mode, a specially designed light
scanner, or projector, is used to project one or more light spots
on the surface of the measured object. Currently, up to 16 spots
in a 4x4 grid can be used (Figure 2).
The measurements space is darkened, so the light spots appear
white in otherwise dark images, and their grey-scale weighted
centroids are easily measured automatically. The intensity of the
light can be adjusted so that the determination of the sub-pixel
position is optimal. The smaller spot size, the smaller details can
be measured. For accuracy reasons, however, the shape of the
spot should be as circular as possible, and the size should be
large enough to give sufficient information for the sub-pixel
position determination. The smallest allowable spot size seems
to be about 3x3 pixels, limiting the smallest detail size to about
0.8 mm. To reach smaller details in the 3-D object space means
that the resolution of the cameras should be increased, or the
cameras should be placed closer to the measured object.
A heuristic algorithm, based on epipolar geometry, is used to
find the corresponding observations among the points seen in
different images. For example, using a 16-point grid, 7-16
points usually get correct corresponding observations from at
least three images. Points having observations in only two
A- 225
tns