ACTA TECHNICA CORVINIENSIS

– Bulletin of Engineering
Tome X [2017]
Fascicule 1 [January – March]
ISSN: 2067 – 3809

1.Miroljub ARBUTINA, 2.Dinu DRAGAN, 3.Srdan MIHIC, 4.Zoran ANISIC

REVIEW OF 3D BODY SCANNING SYSTEMS

1. University of Novi Sad, Faculty of Technical Sciences, Novi Sad, SERBIA
2. DOOB Innovation studio, Novi Sad, SERBIA

Abstract: The purpose of this article is to review 3D body scanning systems currently available in the market. 3D human
scanning is a process of capturing and subsequent digitization of the full body or the body parts into their three-dimensional
graphical representation. As 3D human body scanning techniques are becoming ubiquitous, we give an overview of the
different technologies of 3D body data digitization such as techniques based on photogrammetry, depth sensors, and laser
scanning as well as few more promising techniques. Also, the underlying principles that allow these systems to work are
determined and explained. In order to provide directions for further integration of digitized 3D human bodies, specifications
of 3D body scanning systems are compared, and a variety of commercial scanners available in the market are presented in the
paper.
Keywords: 3D human body scanning, 3D body scanning techniques, 3D body scanning systems, 3D body data digitization, 3D
scanners commercial products

INTRODUCTION techniques appropriate for scanning of humans i.e.

Lately there has been an increased demand for realistic those that commercially available and widespread.
3D models of human beings. Even today there are many These can be divided into three different groups:
indications that it will create a revolution in all aspects Ξ Laser scanning,
of our everyday lives. For example, the way people Ξ Projection of light patterns,
interact (i.e. virtual environments), clothing and styling Ξ Stereo-vision and image processing.
industry, medical field etc. In this light it is obvious why
the demand for 3D scanners of all price ranges has
risen. The main topic of this paper will be to give an
overview of the techniques widely used.
There are many different devices that can be
considered 3D scanners. Any device that measures the
physical world using lasers, lights or x-rays and
generates dense point clouds or polygon meshes can be
considered a 3D scanner. Some common terminology in
usage for them includes 3D digitizers, laser scanners,
white light scanners, industrial CT, LIDAR, and others.
All these devices capture the geometry of physical
objects with hundreds of thousands or millions of
measurements.
Figure 1. Classification of 3D reconstruction methods
All the different approaches to 3D scanning are stated
in Figure1. Some technologies are ideal for short-range As far as reflective ranged measurements are
scanning (<1m, very high detail level), while others are concerned, first distinction is between active and
better for mid (~m)- or long-range scanning (aerial passive methods. With active techniques the light
photography). We will focus on those ranged sources are specially controlled, as part of the strategy

© copyright Faculty of Engineering – Hunedoara, University POLITEHNICA Timisoara

ACTA TECHNICA CORVINIENSIS
– Bulletin of Engineering

to arrive at the 3D information. Active lighting returning to the sensor at, and therefore the distance
incorporates some form of temporal or spatial from the laser source to the object’s surface. Process
modulation of the illumination. This can be a laser light requires precise mechanical apparatus (e.g., by steering
technology or a visible light pattern projective rotating mirrors that reflect the laser light into
technology. With passive techniques, on the other hand, controlled directions), as well as very precise laser
light is not controlled or controlled only with respect to beam. One would also not want the system to take a
image quality. Typically, passive techniques work with long time for scanning. Hence, one ends up with the
whichever reasonable, ambient light available, as is conflicting requirements of guiding the laser spot
often the case in stereo-vision and image processing precisely and fast. These challenging requirements
methods. have an adverse effect on the price. Moreover, total
As this is an overview of relevant 3D acquisition time needed to take one image per projected laser spot
methods, it is not intended to be in-depth nor can be up to seconds or even minutes of overall
exhaustive. The techniques that will be presented are acquisition time. There are faster, special methods,
are widely studied and used for decades. using super-fast imagers, but again at an additional
The organization of the paper is as follows. Section 2 cost.
gives a brief description of the 3D scanning methods In order to remedy this, substantial research has gone
belonging to the three aforementioned groups. Section into replacing the laser spot by more complicated
3 discussed and compares their characteristics. Section patterns. For instance, the laser ray can without much
4 lists some of the commectialy available products.. difficulty be extended to a plane, e.g., by putting a
Section 5 concludes the paper. cylindrical lens in front of the laser. Rather than
OVERVIEW OF 3D SCANNING TECHNIQUES forming a single laser spot on the surface, the
Active range sensors acquire distance measurements intersection of the plane with the surface will form a
from a well-known reference coordinate system to curve (Figure 3). Variant of the scanner with several
points of the object to be reconstructed [1']. They are (usually up to four) scanning pillars have been
very common when highly detailed models are common, without the need to rotate the subject.
required and have been used widely in the industry for Structured Light 3D Scanners
decades. In the past ten years there have been an Structured light scanners also use trigonometric
inrush of low-priced products based on scanning triangulation, but instead of looking at laser light, these
techniques to be described in the following section. systems project a series of linear patterns onto an
Laser Triangulation 3D Scanners object (Figure 4). Then, by examining the edges of each
These scanners use either a laser line or single laser line in the pattern, they calculate the distance from the
point to scan across an object. A sensor picks up the scanner to the object’s surface. Essentially, instead of
laser light that is reflected off the object, and using the camera seeing a laser line, it sees the edge of the
trigonometric triangulation, the system calculates the projected pattern, and calculates the distance similarly.
distance from the object to the scanner(Figure2).

Figure 2. Laser triangulation principle,

image source: Wikipedia
Figure 3. Laser triangulation system
The distance between the laser source and the sensor is
known very precisely, as well as the angle between the Two common methods of stripe pattern generation
laser and the sensor. As the laser light reflects off the have been established: Laser interference and
scanned object, the system can discern what angle it is projection.

18 | F a s c i c u l e 1
 ACTA TECHNICA CORVINIENSIS
– Bulletin of Engineering

The former method allows for the exact and easy Time-of-flight (ToF) scanners
generation of very fine patterns with unlimited depth of Laser pulse-based scanners, also known as time-of-
field. Disadvantages are high cost of implementation flight scanners, are based on a very simple concept: the
and difficulties to provide the ideal beam geometry. The speed of light is known very precisely, so if we know
projection method uses incoherent light and basically how long a laser takes to reach an object and reflect
works like a video projector. A typical measuring back to a sensor, we know how far away that object is
assembly consists of one stripe projector and at least (Figure 5). These systems use circuitry that is accurate
one camera. to picoseconds to measure the time it takes for millions
of laser pulses to return to the sensor, and calculate a
distance. By rotating the laser and sensor (usually via a
mirror), the scanner can scan up to a full 360 degrees
around itself.

Figure 5. Time-of-flight scanning system

Figure 4. Structured light system, image source: Wikipedia
Laser phase-shift systems are another type of time-of-
There are several depth cues contained in the observed flight 3D scanner technology, and conceptually work
stripe patterns. The displacement of any single stripe similarly to pulse-based systems. In addition to pulsing
can directly be converted into 3D coordinates. For this the laser, these systems also modulate the power of the
purpose, the individual stripe has to be identified, laser beam, and the scanner compares the phase of the
which can for example be accomplished by tracing or laser being sent out and then returned to the sensor.
counting stripes (pattern recognition method). Another For reasons that are beyond this paper’s scope, phase
common method projects alternating stripe patterns, shift measurement is more precise.
resulting in binary Gray code sequences identifying the
number of each individual stripe hitting the object. An
important depth cue also results from the varying
stripe widths along the object surface. Stripe width is a
function of the steepness of a surface part, i.e. the first
derivative of the elevation.
In order to get texture extraction, successive
projections of coded and phase-shifted patterns are
required to extract a single depth frame, which leads to
a lower frame rate. Low frame rate means the subject
must remain relatively still during the projection
sequence to avoid blurring, however technology is
maturing fast so this might not be the case in the near
future. The reflected pattern is sensitive to optical Figure 6. Typical scanning setup for either wide-angle ToF
interference from the environment; therefore, or structured-light camera systems
structured-light tends to be better suited for indoor Figure 6 shows a typical scanning setup where the
applications. infrared light wide-angle ToF system is used.
Invisible (or imperceptible) structured light uses Multi-view photogrametry systems
structured light without interfering with other Most ubiquitous method in passive image-based 3D re-
computer vision tasks for which the projected pattern construction systems use the stereoscopic principles as
will be confusing. Example methods include the use of present in human vision. Here, 3D measurements are
infrared light or of extremely high framerates not performed, but 3D information is generated and
alternating between two exact opposite patterns. extracted from a sequence of images, acquired from
However, in those cases it is usually not possible to different viewpoints (Figure 7).
extract texture information.

19 | F a s c i c u l e 1
ACTA TECHNICA CORVINIENSIS
– Bulletin of Engineering

After the matching phase, triangulation process

generates 3D point cloud. Usually, a refining is needed
so a bundle adjustment process is performed [10’]. It is
an optimization procedure that refines an initial
camera and structure parameter estimations for finding
the values that most accurately predict the locations of
the observed points in the set of available images. After
that special interpolation algorithms create dense point
cloud followed by polygonal mesh creation that can be
exported for printing or other usage.
3D SCANNING TECHNIQUES COMPARISON
Figure 7. Multi-view geometry setup Time-of-Flight (ToF) laser systems have an advantage
in precision when measuring distant objects, however
Multi-view geometry is considered an evolution of for small distances the error is usually of the order
stereo-based methods, where only two images were millimeters because of the difficulty in measuring
used: first, correspondences between points in both round-trip time precisely. Triangulation systems have
images were established (matching) and then their limited range but relatively high accuracy of the order
position in 3D space was determined by triangulation. of micrometers. but require a good baseline. A
Prior to acquiring images, several things can be done to disadvantage of ToF systems is that surface texture is
improve reconstruction quality: not captured and that errors will be substantially larger
Ξ Proper coverage of the subjects from all angles to for dark surfaces, which reflect little of the incoming
prevent occlusions, including uniform lightning. signal. Missing texture can be resolved by adding a
Ξ Cameras should be calibrated prior to acquisition of camera, as close as possible to the ToF scanning head.
images to obtain their intrinsic parameters used in But of course, even then the texture is not taken from
triangulation. exactly the same vantage point. The output is typically
Ξ A 3D reference frame defining scale, position and delivered as a massive, unordered point cloud, which
orientation in the form of control points (coded may cause problems for further processing.
targets), lengths of photo-identifiable objects, In contrast to stereo vision or triangulation systems,
and/or angles between photo-identifiable objects. ToF systems are very compact: the illumination is
A major challenge in stereo vision photogrammetry is placed just next to the lens, whereas the other systems
solving the correspondence problem [2’][3'] (Figure 8): need a certain minimum baseline. In contrast to laser
given a point in one image, how to find the same point scanning systems, no mechanical moving parts are
in the other cameras? Until the correspondence can be needed.
established, disparity, and therefore depth, cannot be It is a direct process to extract the distance information
accurately determined. Solving the correspondence out of the output signals of the ToF sensor. As a result,
problem involves complex, computationally intensive this task uses only a small amount of processing power,
algorithms for feature extraction and matching [4’’]. again in contrast to stereo vision, where complex
This area is well studied and there are many mature correlation algorithms are implemented. After the
algorithms [2’][5’] and commercially available software distance data has been extracted, object detection, for
solutions such as Point Cloud Library[6’], Photoscan example, is also a straightforward process to carry out
[7’], Australis [8’] etc. An exhaustive list can be seen because the algorithms are not disturbed by patterns
here [9’]. on the object.
Time-of-flight cameras are able to measure the
distances within a complete scene with a single shot. As
the cameras reach up to 160 frames per second, they
are ideally suited to be used in real-time applications.
High resolution ToF scans, which collect millions of
points, can take more than few seconds which can
result in distortion from motion [11'].
The advantage of structured-light 3D scanners is speed
and precision. Some existing systems are capable of
scanning moving objects in real-time. In triangulation
based systems, the range and depth variation are
limited compared to other methods, but they have a
Figure 8. Multi-view point correspondence problem, great precision. A major advantage of structured-light is
image source: Noah Snavely, Cornell University

20 | F a s c i c u l e 1
 ACTA TECHNICA CORVINIENSIS
– Bulletin of Engineering

that it can achieve relatively high spatial (X-Y) in order to prepare the file for the 3D printer. Scanning
resolution by using off-the-shelf DLP projectors and HD sessions themselves can take several minutes and often
color cameras. must be repeated in order to properly capture the
In image based methods feature extraction and complete geometry or form[14'].
matching also require sufficient intensity and color Many of the modern scanners use these techniques in
variation in the image for robust correlation. This combination. For example, there are cases of scanners
requirement renders stereo vision less effective if the which provide laser scanning precision with
subject lacks these variations i.e. when the local photogrammetric texture extraction component. Also
appearance is uniform within the neighborhood of each as the state of these technologies is in constant
candidate point to be matched. Clothing can have a lack movement it is impossible to make definite conclusion.
of significant local variation in their appearance or Wide range of available commercial 3D scanners can be
present a repeated pattern. This can, however, be reme- seen in [12']. Table 1 shows the empirical comparison
died to some extent by placing artificial fiducial of scanningtechniques, in terms of speed and precision,
markers on the subject. ToF sensing does not have this of the same price category. In Table 1, 5 stars represent
limitation because it does not depend on color or the best empirical value.
texture to measure the distance. Secondly, occlusions in
Table 1. Rough comparison of 3D scanning techniques
the scene make the correspondence between images Scanner type: Speed (5 stars) Precision(5 stars)
ambiguous or even impossible, e.g. articulation of the
Multi-view ***** **
body which leads to self-occlusions.
Triangulation *** *****
In stereo vision, the depth resolution error is a
ToF ** ****
quadratic function of the distance while other two
methods are much better in this respect. The main After creation of the point cloud, by some of the de-
advantage of stereo vision is a very low cost of scribed techniques, discarding of erroneous points (i.e.
implementation as off-the-shelf cameras can be used. outliers) and noise removal by various statistical
The main benefit of 3D laser scanning is that it provides methods can be performed. Some of these outliers can
higher resolution and detail than photogrammetry. be filtered by performing a statistical analysis on each
Therefore, facial details and even details on clothing are point's neighborhood, and trimming those which do not
potentially captured more intricately with a 3D meet a certain criterion. The sparse outlier removal
scanner, making the resulting 3D printed figurine more implementation in the publicly available Point Cloud
detailed and realistic as well. Scanning can be Library is based on the computation of the distribution
accomplished in a scanning booth, most often found in of point to neighbors distances in the input data-set.
retail locations, or by using a 3D scanner in virtually For each point, the mean distance from it to all its
any type of indoor or outdoor setting. A number of 3D neighbors is computed. By assuming that the resulted
laser scanners are available as portable handheld distribution is Gaussian with a mean and a standard
devices, offering quick and easy set up and flexibility in deviation, all points whose mean distances are outside
the location of the scan session. Structured light an interval defined by the global distances mean and
scanners can offer greater speed and accuracy than standard deviation can be considered as outliers and
laser scanners when setup with large fields of view, trimmed from the data-set.
approaching that of photogrammetry, but usually
require several scanners to be synced together in order
to capture the required angles without any
“shadowing” due to “line of sight” and binocular vision
restrictions. These multi-sensor systems, while very
effective once set up, can be difficult and sensitive to
calibrate.
The downside of using laser scanners for body form
applications is that they take a second or more to
capture data on any part of a subject during a scan.
While that might not seem like much, the living body
will move ever so imperceptibly during that second.
Even if the movement is only a millimeter, the result
will be a double layer of scan data in the same area of
the geometry, which will cause problems when creating
a mesh. Labour-intensive and often complex
adjustment of the scan data is required after scanning Figure 9. Typical 3D mesh model

21 | F a s c i c u l e 1
ACTA TECHNICA CORVINIENSIS
– Bulletin of Engineering

Then a dense point cloud is generated using High-end, metrology-grade, portable scanners like
appropriate interpolation algorithm followed by those by Artec 3D (Eva and Space Spider) and Nikon
filtering of the points from the result to adjust it and (ModelMaker MMCx on Fig. 12) offer less than 0.5mm
eliminate bad points. After that, creating polygon mesh resolution, depending on the scanning distance. They
model can be done for the model to be used in other 3D are rarely used in full body scanning as the process can
modeling software. An example of a generated mesh take minutes and the obtained level of details is
model without added texture is shown below (Fig. 9). excessive for non-special purposes.
OVERVIEW OF COMMERCIAL PRODUCTS
There are numerous commercial scanners on the
market today, with the number and variety constantly
increasing. Hand-held 3D scanners are gaining
popularity due to the ease of use and portability. Their
pricesrange from 100 euros up to 20000 euros.
Some commercial solutions and the techniques they
use:
Ξ Microsoft Kinect 1 uses a pattern of projected
infrared points to generate a dense 3D image.
Ξ Microsoft Kinect for Xbox One uses a wide-angle
time-of-flight camera.
Ξ Intel RealSense camera (Fig. 10 ) projects a series of
infrared patterns to obtain the 3D structure. Figure 12. Nikon ModelMaker MMCx 3D laser scanner
Low cost models like Intel RealSense 3D cameras are Structured light scanners also havewide price
quickly finding their way in people’s everyday lives. spectrum. In the price range of 3000euros structured-
They are becoming a part of notebooks, phones and light scanners products have superiority. An example is
handheld devices, replacing the old 2D cameras, or as shown on Fig. 13. Submilimeter resolution and high
an add-on accesory These devices are mainly used for mesh density for scans that last only several secons has
the scanning of parts of human body (facial made them videly used. For full body scanning, either a
expressions, hand gestures, etc.) for entertainment turntable has to be used or a set-up of several scanners.
purposes as their resolution and texture detail tends to
be low.

Figure 10. Intel RealSense 3D cameras and their place

Figure 13. David SLS-3-STEREO structured light scanner
on the smartphones
An emerging technology is the real time 3D scanning
capability. It has applications in animation and medical
imaging to capture real facial expressions and body
shape changes in real time. Documenting and
quantifying anatomical dense surface movement, pose,
and expression. 3D camcoders have been on market for
some time, and are becoming more affordable. In case
of real time 3D laser scanner they are still not present
on the market
CONCLUSIONS
Active sensors provide directly the range data
containing the 3D coordinates necessary for the mesh
Figure 11. iSense for iPads by 3DSYSTEMS generation phase. Active range scanning technology has

22 | F a s c i c u l e 1
 ACTA TECHNICA CORVINIENSIS
– Bulletin of Engineering

been applied to automatically acquire highly accurate References

geometric data of people. However, generally range [1.] T.C. Azevedo, “3D Object Reconstruction using
based scanners can only capture a single static pose of a Computer Vision: Reconstruction and
person. Passive sensors provide images that need a Characterization Applications for External
mathematical model to derive the 3D coordinates. After Human Anatomical Structures”, PhD Thesis,
the measurements, the data must be structured and a Faculdade de Engenharia da Universidade do
consistent polygonal surface generated to build a Porto, pp. 7-32.
realistic representation of the modeled objects or [2.] S. Gold, A. Rangarajan, C.P. Lu, S. Pappu, E.
scenes. A photo-realistic visualization can afterward be Mjolsness, “New algorithms for 2D and 3D point
generated by texturing the 3D models with image infor- matching: Pose estimation and correspondence”,
mation. Pattern recognition, vol. 31, no. 8, pp. 1019-1031
Nowadays, when choosing appropriate scanner suiting 1998.
ones’ needs, a wide range of products currently [3.] T. Moons, L.V. Gool, M. Vergauwen, “3D Re-
available is rapidly increasing. The old classification of construction from Multiple Images Part 1:
laser scanners being the most expensive is no longer Principles”, Foundations and Trends in Computer
true. Same goes for structured-light scanners. Both Graphics and Vision, vol. 4, no. 4, pp. 287–398,
methods have products of full body millimeter 2008.
precision scanners for 2000$ or less [12']. Also small [4.] D. Dragan, S. Mihic, Z. Anisic, I. Lukovic, “Role of
and portable scanners are finding their way to the background subtraction in creating human body
market, being pushed by large companies with point clouds from photos”, The 6thInternational
aggressive pricing of several hundred dollars. If one Conference and Exhibition on 3D Body Scanning
needs a high quality texture model currently the choice Technologies, 27-28 October, 2015, Lugano,
boils down to multi-view photogrammetry scanners. Switzerland, pp. 210-217.
However, structured-light scanners might not be too far [5.] M.M. El-gayar and H. Soliman, “A comparative
behind. study of image low level feature extraction
The race is on among manufacturers of digital image algorithms”, Egyptian Informatics Journal, vol. 14,
and geometry capture technology to develop a solution no. 2, pp.175-181, 2013.
that is affordable, off the shelf, quick and easy to set up, [6.] [6] R.B. Rusu and S. Cousins, Willow Garage,
high resolution and fast, in order to accommodate “3D is here: Point Cloud Library (PCL)”, The 2011
movement. To this end, photogrammetry solution IEEE International Conference on Robotics and
providers are trying to incorporate multi-camera Automation (ICRA), pp. 1-4.
systems and software into one simple, packaged [7.] Agisoft, “Photoscan”. Available at:
solution. At the same time, makers of 3D laser scanners http://www.agisoft.com/[Accessed 20June
are trying to develop faster and more automatic 2016].
geometry capture solutions to accommodate for [8.] Photometrix, “Austrlis. Available
movement and shorter scanning sessions. There are at:http://www.photometrix.com.au/australis/[A
also major steps being taken by mainstream electronics ccessed 20June 2016].
companies and suppliers to incorporate 3D content [9.] “Comparison of photogrammetry software”,
capture into mobile phones or tablets, and this is where Wikipedia. Available at
we are already seeing a big change in the ability of https://en.wikipedia.org/wiki/Comparison_of_p
average consumers to generate usable 3D content. hotogrammetry_software[Accessed 20June
As the need for 3D models of humans will certainly rise 2016].
in the near future, the market of 3D scanners is [10.] B. Triggs, P. McLauchlan, R. Hartley, A. Fitzgibbon,
becoming a lucrative business and the competition is “Bundle Adjustment — A Modern Synthesis”, in
developing new and innovative solutions at a International workshop on vision algorithms,
staggering rate. It is certainly a field where a potential 1999. Springer Berlin Heidelberg. pp. 298-372.
buyer must be up-to-date on a monthly basis. [11.] L. Li, “Time-of-Flight Camera – An Introduction”,
Note Technical White Paper SLOA190B – January 2014
This paper is based on the paper presented at The 7th Revised May 2014, pp. 1-6. Available at:
International Conference on Mass Customization and http://css1a0.engr.ccny.cuny.edu/~wolberg/cap
Personalization in Central Europe – MCP–CE 2016 – stone/kinect/ToFBasicsTI14.pdf[Accessed
Mass Customization and Open Innovation, organized in 20June 2016].
Novi Sad, SERBIA, September 21-23, 2016, referred [12.] Kira, “The Best 3D Scanners of 2015”,
here as [15]. www.3ders.org, 2015. Available at:

23 | F a s c i c u l e 1
ACTA TECHNICA CORVINIENSIS
– Bulletin of Engineering

http://www.3ders.org/articles/20151209-best-
3d-scanners-2015.html[Accessed 20June 2016].
[13.] Ν. D’Apuzo, “State of the art of the methods for
static 3D scanning of partial or full human body “,
Proceedings of Conference on 3D Modelling, June
13-14, 2006, Paris, France, pp. 1-8.
[14.] Julie Reece, “The Pros and Cons of
Photogrammetry vs. 3D Scanning for 3D Printed
Figurines”, Available at:
http://mcortechnologies.com/the-pros-and-
cons-of-photogrammetry-vs-3d-scanning-for-3d-
printed-figurines-blog/[Accessed 20June 2016].
[15.] Miroljub Arbutina, Dinu Dragan, Srdan Mihic,
Zoran Anisic, Review of 3D body scanning
systems, The 7th International Conference on
Mass Customization and Personalization in
Central Europe – MCP–CE 2016 – Mass
Customization and Open Innovation, Novi Sad,
SERBIA, 2016

ISSN:2067-3809
copyright © University POLITEHNICA Timisoara,
Faculty of Engineering Hunedoara,
5, Revolutiei, 331128, Hunedoara, ROMANIA
http://acta.fih.upt.ro

24 | F a s c i c u l e 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.