Q-switched or pulsed
ToF
laser systems (Zubrin
et al.
1991; Wil-
son
et al.
1999; Zayhowski 2010, 2018). For example, passive
Q-switching of a laser was incorporated into Cyra Technologies
laser scanners via John Zayhowski of the Massachusetts Insti-
tute of Technology (
MIT
) Lincoln Laboratory—a research labora-
tory with a long history of developing laser systems for space
and defense projects (Jelalian 1992; Delaney and Ward 2000;
Gschwendtner and Keicher 2000; Grometstein 2011). Zayhows-
ki recognized that a 532-nm green laser using a Q-switched
system would fit the specifications given to him by Ben Kacyra
and Jerry Dimsdale (seen in Figure 5). It addressed eye-safety is-
sues that could occur in a commercial-based laser scanning unit
(Wilson
et al.
1999; Zayhowski 2010). The diameter of the laser
beam and the use of Q-switching provided a means of adjusting
the intensity of a laser that was otherwise ideal for survey-based
applications. Other requirements needed for the Cyra Technolo-
gies based midrange TLS systems—such as the timing circuit
that came out of the Department of Energy's Los Alamos Na-
tional Laboratory—are discussed in more detail in part two of
this article. The incentive to commercialize technologies linked
to the
SEI
, which included the
JPL
rover seen in Figure 1a, came
about when manned missions to Mars were abandoned for
cheaper, robot-based solutions in 1992 (Hogan 2007).
Policy based influences on turning point moments to come
out of the US can be traced back to the Stevenson-Wydler
Technology Innovation Act of 1980. This started a chain of
legislation passed over the next two decades, which enabled
federal laboratories to transfer technologies to non federal
entities in the US. The last of note in this time period being
the Small Business Technology Transfer Act of 1992.
Best-Fit Solution and Developments in Computing
Midrange
TLS
is shaped by end-user requirements, the ap-
plications around which services or solutions are developed,
and trends linked to computer processing. These contributing
factors stretch across all phases of its history, be it space-based
applications, machine vision for use in autonomous vehicles,
or use for “as-built” surveying (explored in more detail in part
two of this article). Its transition to industrial uses for “as-
built” information collection was also stimulated by the need
for safer forms of environmental data collection in environ-
ments otherwise hazardous to humans. For example, two of the
earliest European and North American companies to seriously
explore midrange
TLS
outside of research applications did so
for nuclear and industrial plant applications (Addison and
Gaiani 2000; Kacyra
et al.
1997; Pot, Thibault and Levesque
1997). Cyra Technologies had stemmed from observations
made by Ben Kacyra when he worked in nuclear and industrial
plants at his engineering company, Cygna (Cheves 2014). Ben
envisioned what was later described to the
MIT
Lincoln Labora-
tory (by Cyra Technologies) as a solution that could produce
something like a “3D Polaroid” of the scene collected. On the
other hand, triangulation-based Mensi systems were designed
to meet the accuracy, repeatability, and resolution requirements
needed to document nuclear power plants owned and man-
aged by Électricité de France (
EDF
) (Pot
et al.
1997).
In its commercial period of use, midrange
TLS
became an
answer to problems associated with preexisting computer-
aided design (
CAD
) workflows (Pot
et al.
1997). This included
idealized representations or designs in working environments
where not knowing the imperfections in a built environment
could cost lives. Midrange
TLS
was able to easily document an
object or scene, enabling
CAD
drawings or plans to be based
on real-time conditions at a level of detail not seen before (Pot
et al.
1997; Zheng, Lewis and Gethin 1996). Prior to this,
CAD
-
based models ran the risk of being detached from their real-
world counterparts. The resolution of information capture
was restricted to collecting a series of points on a surface, as
opposed to collecting a digital mold of the surface itself. This
was especially the case in environments where everything
had known measurements assigned to it, such as pipe-filled
spaces in industrial plants.
Enter Mensi
The first Mensi
SOISIC
scanners were developed around opti-
cal triangulation because of the accuracy, repeatability and
resolution requirements of their application in industrial
power plants in the early 1990s (X. N. Chen
et al.
2005; Shan
and Toth 2008). There the driving force was the speed of data
recording in relation to safer working conditions for the end
user. A system architecture based around optical triangula-
tion was adopted due to the restrictions associated with
PS
and
ToF
systems of the time (X. N. Chen, email to author,
October 18, 2013; Fienup 2013). The accuracies and ranges at
which information could be collected were not yet to mil-
limeter standard (Pot
et al.
1997), and computer-based timing
and calculation were not yet powerful enough to make
ToF
and
PS
viable options for high-resolution documentation and
measurement (Wilson
et al.
1999; Fienup 2013). It was only
later, through improvements in integrated timing circuits and
algorithm-based noise filtering in the waveform of the laser
beam, that
PS
and
ToF
became the standard system architec-
tures for midrange
TLS
(Hebert and Krotkov 1992; Froehlich
1997; Flatscher
et al.
1999; Wilson
et al.
1999; Fienup 2013).
It is no coincidence that Mensi replaced optical triangulation
completely by 2001, with the GS 100 scanner bringing their
technology more in line with Cyra Technologies and Riegl
systems. (X. N. Chen
et al.
2005; Shan and Toth 2008). By
this time, computer graphics-based processing was becom-
ing powerful enough to make working with point-cloud data
easier and more affordable. 3D graphic architectures like
RealityEngine from Silicon Graphics (
SGI
) had evolved into
open standard application programming interfaces, such as
OpenGL, which also made applications portable between
devices (Akeley 1993). In other words, it was now easier for
developers to create commercial point cloud-based software
because of the continuity such an open standard provided–
much in the same way the MIDI file format did for sound
engineers and musicians (Loy 1985).
Point Cloud-Based Computing Before Graphics Cards
Earlier examples of midrange
TLS
, like the
ALV
project, high-
light restrictions caused by computing power available at
the time of the Strategic Computing Initiative (Roland and
Shiman 2002). Onboard parallel computing—which enabled
tasks or complex sets of information to be separated into
smaller calculations and processed simultaneously—had to be
used in order to work with terrain data in real time (Weems
et al.
1991; Chuck Thorpe, email to author, Month DD, 2015).
It was expensive but necessary due to the fact that the only
other hardware capable of doing this was mainframe comput-
ers. Parallel processing allowed for the custom manufacture
of a computer small enough and light enough to fit inside the
ALV
(Weems
et al.
1991; Roland and Shiman 2002).
Point cloud based information became more accessible to
a broader population of people because of graphic processing
units (
GPU
). These started to make graphic intensive computer
based tasks more affordable to a general consumer by the end
of the 1990s. Prior to this, graphic based work carried out on
most micro computers (an early term for a personal computer)
were handled by the central processing unit and custom
chipsets. Even by the mid-1990s, point cloud-based software
like Computer Graphics Perception (
CGP
) was developed on
specialist SGI and Sun Microsystems hardware—even though
softwares like
CGP
were designed to run on standard Windows
NT machines and Intel 486 processor-based laptops of the time
(Kacyra
et al.
1997). Software and hardware development was,
424
July 2020
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
