Professional Documents
Culture Documents
Triangulation Based Laser Scanner
Triangulation Based Laser Scanner
During the recent times, the need for contact free measurement for mathematical
representation of physical objects has been growing ranging from transferring
modeled clay or wood to a CAD model to analyzing competitors products, with
recent applications extending to the animation industry. But often they still
require complex and expensive equipment. Such equipment unfortunately is
too expensive to be afforded by the general public. Access to such scanning
techniques could prove to be a great boon to students and hobbyists whose
interests may concern reverse engineering a particular component or in modeling
a particular object.
With the increasing efficiency of hardware and software devices, alternative
low cost solutions are in great demand. Such solutions should not only be cheap,
they should also present a relative ease and simplicity in their assembly. In this
report, we propose such a low-cost system for 3d data acquisition and surface
registration. The scanner uses a reference surface that acts as the coordinate
system of the object. The object placed in front of the reference surface also acts
as a system of the reference surface, and both these surfaces, automatically come
together. A single-view image is obtained by sweeping a laser line over the object
while keeping the camera fixed and analyzing the acquired laser stripes. The line
and the camera can move independently, making it possible to conveniently scan
an object while viewing it from different views. The only hardware requirements
are a simple handheld point LASER, a glass rod (cylindrical lens) and a standard
camera.
Contents
1 Introduction 3
2 Structural Design 5
3 Experimental Procedure 7
3.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2 Initial Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3 The Scanning Process . . . . . . . . . . . . . . . . . . . . . . . . 8
3.4 Shape Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4 Algorithms 10
1
List of Figures
1.1 Intersection of the illuminating laser beam and the rays projected
back . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2
Chapter 1
Introduction
Laser scanning has become widely popular with the advent of Reverse Engineer-
ing techniques in the field of manufacturing and rapid prototyping techniques.
Scanners which measure co-ordinates of physical objects by triangulating dif-
ferent light beams are well known since more than the past two decades. The
scanning process could be done using a myriad of techniques like structured
light, coded light, time of flight, etc. Laser range scanners are commonly used for
contactless measuring of surfaces and 3d scenes in a wide range of applications.
The field of application includes, areas like manufacturing, reverse engineering,
robotics, industrial design, medical diagnosis, archaeology, multimedia and web
design, as well as rapid prototyping and computer-aided quality control.
Most commercial laser scan systems use a camera and a laser beam or laser
plane. The measurement here is based on triangulation, i.e. the intersection of
the illuminating laser beam and the rays projected back to the camera as shown
in the Figure 1.
This is generally coupled with expensive high precision actuation, for ro-
tating/translating the laser plane the object. With recent advancements in
computational techniques, devices have been developed which avoid the use of
expensive actuators. These approaches have to determine the position and orien-
tation of the laser device on-line and to permanently avoid an external tracking
system. In place of an external tracking system, we propose a real-time self-
calibration of a hand-held laser plane, which is based on a simple analysis of the
laser stripes in the camera images. Thus, the laser line can be swept manually
over the object during the scan. This method has the following advantages:
1. Low cost
2. Flexibility
3. Simplicity in Design
This approach has been investigated by Lyubomir Zagorchev and Ardeshir
Goshtasby of the Wright state University[4]. They use a reference double-frame,
which is placed around the object and acts as calibration target. The approach,
involves calibration of the LASER using the four visual intersection points of the
laser and the double-frame. We could however make the calibration more flexible
by designing the process in such a way that the calibration target can be almost
3
Figure 1.1: Intersection of the illuminating laser beam and the rays projected
back
4
Chapter 2
Structural Design
The hardware organization of the scanner is depicted in Figure 2.1. The basic
design of this device is quite simple. The laser ray, expanded to a plane by a
cylindrical lens, has to intersect two things at the same time: the (unknown)
surface, and the a priori known reference geometry (usually the background).
The visible intersection with the background is used to calibrate the laser (cal-
culate the spatial orientation of the LASER plane). With this knowledge we
can triangulate new 3d point coordinates of the objects surface by intersecting
the laser plane with the projecting rays. Certainly, the camera must have been
calibrated so that its external and internal parameters are exactly known. Thus,
the exact coordinates of the background structure with respect to the camera
coordinate system are implicitly known.
As mentioned earlier, Scanning is done by sweeping the laser light over an
object by hand. In this manner, dense range data can be captured in some areas
while sparse data are captured in other areas. Effectively, the laser sweeping
process can be automated using a rotating mirror, a rotating arm, or a transla-
tion stage. The global reference for the computation is attached to the reference
surface, and as long as the position of the object remains fixed with respect to
the surface, object coordinates determined from different views will all be in
5
the same coordinate system and thus will automatically produce a unique set
of data pertaining to the object. The scanner, therefore, makes it possible to
combine different-view range images of an object without registering them.
6
Chapter 3
Experimental Procedure
7
Figure 3.1: Reference plane whose dimensions and coordinates are known.
8
Figure 3.2: Scanning in process
9
Chapter 4
Algorithms
p = r ∩ ELaser (4.2)
where,
p is the surface plane of the object under scan.
r is the line laser image
10
Figure 4.1: Laser triangulation
11
Figure 4.2: Surface Registration of two surfaces A and B
with,
distB (x) = minkx −A TB (4.6)
y∈B
Then Ω, the proportion of overlapping area can be computed with 95% accuracy
as: Pn
contactB (xi ) 1.96
Ω = i=1 ± √ (4.7)
n 2 n
12
Chapter 5
3. Explore the possibilities of converting the obtained scan from .obj file to
STL or VRML formats which could then be uploaded to a prototyping
machine.
13
4. Include support sharp edges using white light scanning.
The hardware and computation involved in the design and working of the
LASER scanner were discussed. The scanner uses a reference surface to capture
images and computationally merge them.
Thus, we demonstrated a very effective but simple LASER scanner, with a
decent accuracy for most of the common purposes and sufficient to build up a
system for 360-object-reconstruction, which is on par with the other commer-
cially available, expensive techniques.
14
Bibliography
[1] Besl, P., and McKay, N. A method for registration of 3-d shapes. IEEE
Transactions on Pattern Analysis and Machine Intelligence 14, 2 (1992),
239–256.
[2] Fischler, M. A., and Bolles, R. C. Random sample consensus: a
paradigm for model fitting with applications to image analysis and auto-
mated cartography. Commun. ACM 24, 6 (1981), 381–395.
[3] Simon, W., Sven, M., and Friedrich, M., W. Pattern Recognition.
Springer Berlin / Heidelberg, 2006.
[4] Zagorchev, L., and Goshtasby, A. A paintbrush laser range scanner.
Computer Vision and Image Understanding 101, 2 (2006), 65 – 86.
15