Evaluation of Low-Cost 3D Scanner Hardware for Clothing Industry

Michael Danner

1 a

, Elena Alida Brake

1 b

, Christian Decker

1 c

, Matthias R

atsch

1 d

Yordan Kyosev

2 e

and Katerina Rose

1 f

Reutlingen Research Institute, Reutlingen University, Alteburgstraße 150, Reutlingen, Germany

Institute of Textile Machinery and High-Performance Material Technology,

Chair of Development and Assembly of Textile Products, Technische Universit

at Dresden, Dresden, Germany

{michael.danner, elena.brake, christian.decker, matthias.raetsch, katerina.rose}@reutlingen-university.de,

Keywords:

3D Body Scanner, 2D-3D Reconstruction, Textile Technology, Garment Fit, Mental Health.

Abstract:

In recent years, the demand for accurate and efﬁcient 3D body scanning technologies has increased, driven

by the growing interest in personalised textile development and health care. This position paper presents

the implementation of a novel 3D body scanner that integrates multiple RGB cameras and image stitching

techniques to generate detailed point clouds and 3D mesh models. Our system signiﬁcantly enhances the

scanning process, achieving higher resolution and ﬁdelity while reducing the cost, time and effort required for

data acquisition and processing. Furthermore, we evaluate the potential use cases and applications of our 3D

body scanner, focusing on the textile technology and health sectors. In textile development, the 3D scanner

contributes to bespoke clothing production, allowing designers to construct made-to-measure garments, thus

minimising waste and enhancing customer satisfaction through ﬁtting clothing. In mental health care, the 3D

body scanner can be employed as a tool for body image analysis, providing valuable insights into the psy-

chological and emotional aspects of self-perception. By exploring the synergy between the 3D body scanner

and these ﬁelds, we aim to foster interdisciplinary collaborations that drive advancements in personalisation,

sustainability, and well-being.

1 INTRODUCTION

3D body scanning technology has gained signiﬁcant

attention in recent years, revolutionising various in-

dustries by providing a means for accurate, non-

invasive, and efﬁcient body measurements. The abil-

ity to capture detailed, three-dimensional data of the

human body in mere seconds has paved the way for

countless applications, ranging from healthcare and

ﬁtness to the textile industry and performance diag-

nostics in sports and entertainment (Schlich et al.,

2010).

Traditionally, measuring the human body has been

a time-consuming and error-prone process involv-

ing manual measurements and subjective evaluations.

The advent of 3D body scanning has not only stream-

https://orcid.org/0000-0002-8652-6905

https://orcid.org/0009-0004-7828-6501

https://orcid.org/0009-0003-6030-5498

https://orcid.org/0000-0002-8254-8293

https://orcid.org/0000-0003-3376-1423

https://orcid.org/0000-0003-4294-9700

lined this process but also enhanced the precision and

consistency of the data collected. By employing tech-

niques such as structured light, laser triangulation, or

photogrammetry, 3D body scanners can create accu-

rate digital representations of the human form, cap-

turing intricate contours and dimensions (Daanen and

Psikuta, 2018).

The versatility of 3D body scanning has allowed

for its implementation in various sectors. In health-

care, it can facilitate diagnoses, monitor progress, and

assist in rehabilitation. In ﬁtness, it provides valuable

insights into body composition and tracking physical

changes, for example, in the analysis and diagnosis of

high-performance athletes. The textile industry bene-

ﬁts from made-to-measure clothing pattern construc-

tion and improved sizing systems, while the entertain-

ment sector can utilise the technology for character

modelling, virtual reality, and motion capture anima-

tions capturing (Rozmus et al., 2021).

As we continue to explore the capabilities of 3D

body scanning technology, it is crucial to foster di-

alogue and collaboration among researchers, practi-

Danner, M., Brake, E., Decker, C., Rätsch, M., Kyosev, Y. and Rose, K.

Evaluation of Low-Cost 3D Scanner Hardware for Clothing Industry.

DOI: 10.5220/0012231800003543

In Proceedings of the 20th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2023) - Volume 1, pages 727-735

ISBN: 978-989-758-670-5; ISSN: 2184-2809

727

tioners, and industry stakeholders to address chal-

lenges, share advancements, and ultimately unlock

the full potential of this transformative technology.

This paper presents both the design and construction

of the built 3D scanner’s hard- and software as well

as the post-processing of the collected data to a 3D

avatar representation of the individual, termed as a

’scanatar’. It also includes validation of the scan-

ner’s output by comparison with existing 3D scans

and an evaluation of the scanner’s accuracy and pre-

cision against a deﬁned list of requirements.

2 RELATED WORK

2.1 Low-Cost 3D Scanner

In the area of low-cost 3D scanning, (Iwayama, 2019)

presented a novel system that uses the Raspberry Pi

Zero W for full-body 3D scanning. Speciﬁcally de-

signed for the production of avatars in VRM format,

this system demonstrates the potential of affordable

hardware for creating detailed 3D models. Iwayama’s

approach offers valuable insights into our work, par-

ticularly in terms of cost-effective hardware usage.

Garsthagen presented a low-cost, open-source,

multi-camera, whole-body 3D scanner. It was used in

a wide range of applications and proved its versatility

and effectiveness (Garsthagen, 2014).

The research of Zeraatkar and Khalili (Zeraatkar

and Khalili, 2020) details the creation of a cost-

effective and speedy human 3D scanner, highlighting

its design, construction process, and potential appli-

cations in diverse ﬁelds, notably in areas where tradi-

tional high-cost technologies are less prevalent.

In the investigation of Kyosev and Siegmund

(Kyosev and Siegmund, 2019) the Asus RGB-D cam-

era was found to be accurate enough for the apparel

industry, with a mesh size of 3-4 mm for scanning a

human body. However, challenges were encountered

due to optical systems and body motion, resulting in

mesh discontinuities. An algorithm was developed to

evaluate gaps in the clothing based on the generated

meshes, but further research is required to overcome

limitations and achieve full automation in measuring

clothing gaps.

2.2 Scanner Operating Modes

The operational methodologies employed by low-cost

3D scanning devices can be classiﬁed into three pri-

mary techniques, namely Photogrammetry, Triangu-

lation, and Structured Light:

• Photogrammetry is a cost-effective 3D scanning

method that measures object distances using pho-

tos. It offers precise data and diverse uses. This

technique captures multiple photos from various

angles and employs software to reconstruct a 3D

model. (Journalists.org, 2020; Vedantu, 2023)

Photogrammetry is utilised in mapping, archi-

tecture, engineering, manufacturing, quality con-

trol, forensics, and law enforcement. Schenk’s

(Schenk, 2005) introduction covers photogram-

metry’s principles, applications, advancements,

challenges, limitations, and potential solutions.

• Triangulation is another technique used in 3D

scanning. It involves projecting a laser point or

line onto an object and then using a camera posi-

tioned at a known distance from the laser source to

capture the location of the laser on the object. The

distance to the object is then calculated using the

principles of triangulation(Franc¸a et al., 2005).

• Structured light projects a known pixel pattern

onto an object, allowing vision systems to cal-

culate depth and surface information. It is com-

monly used in applications like facial recognition

and body scanning. (Rocchini et al., 2001)

2.3 Areas of Application

The utilisation of 3D scanners is widespread, with

a multitude of forms and designs available, ranging

from mobile to stationary. The applications of 3D

scanners are equally diverse, with a broad range of ar-

eas in which they are employed. In various industries

and industrial manufacturing, 3D scanners are utilised

to measure, digitise, and analyse the forms and di-

mensions of construction parts, components, or other

objects.

In the realm of art and culture, 3D scans are em-

ployed to digitise art and cultural objects or recon-

struct them using the acquired data. This is done

for archiving and conservation purposes, as well as

to create virtual exhibitions or enable research with-

out the need for a physical object on site. However,

portable hand-held scanners are typically used for this

purpose, as a high level of accuracy is of utmost im-

portance (Akca et al., 2007).

In architecture and construction, the use of 3D

scanners is also common practice to create 3D mod-

els of the object and premises. This is mainly used for

precise and fast surveying for planning and visualisa-

tion in the building industry. In addition, in the ﬁeld

of mapping and surveying, terrain, infrastructure and

environments can also be recorded to a certain extent

(Sepasgozar et al., 2016).

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

728

In healthcare, 3D scanners are used for more pre-

cise detection of body parts or organs for the diag-

nosis and treatment of diseases for medical imaging

(Hitomi et al., 2015). Likewise in the medical ﬁeld,

3D scanners are also used in forensics, where they can

be used to document crime scenes and capture exact

3D models of evidence (Haleem and Javaid, 2019).

In the application area of gaming and entertain-

ment, 3D scanning technology is used to create digi-

tal 3D models of real objects or people and integrate

them into virtual worlds or for motion capturing (Ya-

hav et al., 2007).

In the textile industry, 3D scanners are used in the

design and development of clothing as well as furni-

ture, accessories and seats in the automotive indus-

try. On the one hand, physical objects can be scanned

and digitised so that the dimensions and shapes can be

used in design and prototyping software, such as the

dimensions of a scanned human body. On the other

hand, objects such as car seats, furniture or even worn

clothing can be scanned so the 2D patterns can be ﬂat-

tened from the 3D scans. (D’Apuzzo, 2007) Further-

more, body scanners can be used in this area of ap-

plication to record dynamic body dimensions, such as

the length of the back, which changes when a person

bends forward (Chi and Kennon, 2006). 3D scans are

also used to record the inﬂuence of clothing on the

body, such as changes in the shape of the human soft

tissue caused by compressive clothing or the lifting of

the breast by wearing a bra (Brake et al., 2022).

2.4 Textile Application

The use of 3D scanning technology extends to the ar-

eas of development and prototyping in clothing tech-

nology. By scanning a physical object, developers can

capture a digital model and use it in CAD systems,

which can be modiﬁed and optimised before produc-

tion. This approach not only saves time and resources

but also allows for greater creativity and experimenta-

tion in the design process. Besides scanning garments

to ﬂatten a 2D pattern from the 3D object, 3D scan-

ners are mainly used to scan people and objects to be

measured. Capturing measurements and dimensions

via scanning allows for an error-free method and also

for ﬁt checks in CAD to be done digitally on the avatar

or object (

Spelic, 2020).

In addition, 3D scanners can serve as a valuable

tool for the restoration and conservation of historic

textiles and artefacts. By creating a digital replica

through scanning, experts can examine and analyse

the object without damaging the original. This tech-

nology also enables the production of replicas for ed-

ucational and exhibition purposes of historical textiles

(

Zyła et al., 2021).

The use of 3D scanning technology in the textile

industry has led to signiﬁcant improvements in qual-

ity control. This technology is able to detect even the

most minor ﬂaws or inconsistencies in textile prod-

ucts, ensuring that the ﬁnal product meets the highest

quality standards with a great ﬁt. This beneﬁts not

only the consumer but also the manufacturer, as less

waste is produced and the efﬁciency of the production

process is improved (Jhanji, 2018).

2.5 Contribution

Our goal is to democratise access to 3D scanning

for broad user groups who do not have ﬁnancial re-

sources for a commercial system or who do not have

the know-how to design hardware and software for a

scanner device by themselves. This is achieved by

reducing the complexity, mainly through the use of

standards and adherence to best practices that work in

the environments of those user groups.

The full-body 3D scanner was built using com-

mercial off-the-shelf (COTS) cameras and computing

hardware. The software then allows different parts to

act like a single device. This paper presents the de-

sign and construction process and the performance of

the built 3D scanner but with only 48 cameras. It also

includes validation and evaluation of the scanner’s ac-

curacy and precision against a deﬁned list of require-

ments. The main problem addressed in this study is

the need for a cost-effective yet accurate 3D scanner

that can be used in the ﬁeld of apparel engineering by

researchers who are not software experts.

Figure 1 shows our installation of the 3D scan-

ner. It consists of 48 Raspberry Pi 4 model B (RPi,

2023a) computers mounted in a frame around a 2m x

2m ground plate. Each Raspberry Pi connects to an

8 Megapixel camera module v2.1 (RPi, 2023b). The

subject stands in the middle of the ground plate, and

all surrounding cameras capture pictures simultane-

ously. These pictures are downloaded by a user for

post-processing, stitching all single pictures together

to create a 3D image.

The scanning software project is open source and

available on Github

2.6 Deﬁnition of Requirements

From the previous deﬁnitions of the areas of applica-

tion and possible options for using the scanner, it is

important to precisely deﬁne the requirements to be

met by the scanner in order to use it in these speciﬁc

cases.

https://github.com/cdeck3r/3DScanner

Evaluation of Low-Cost 3D Scanner Hardware for Clothing Industry

729

Figure 1: 3D Scanner installation.

• Capturing body measurements: The scanner must

be able to capture precise measurements of body

dimensions such as lengths, girths and circumfer-

ences (Bartol et al., 2021). This is important to

construct and produce customised garments.

• Capture of body forms: The scanner should also

be able to capture the whole individual body

shapes and proportions of each scanned indi-

vidual, including special physical characteristics

such as asymmetries. This is important to ensure

a correct ﬁt, taking into account individual physi-

cal conditions (Ashdown et al., 2004).

• Speed: The scanner should capture data as

quickly as possible in under 1 second and be efﬁ-

cient to save time and optimise workﬂow.

• Accuracy: The scanner should be highly accu-

rate to ensure that the captured data is reliable

and allows for precise measurement of the body.

The maximum deviation of the scan from the real

dimensions of the body should be within 5mm

(

Spelic, 2020).

• Compatibility: The scanner should be compatible

with various software and hardware systems to en-

sure smooth integration into existing workﬂows.

The above requirements were identiﬁed through an

analysis of the speciﬁc needs and challenges of the

apparel industry (Kyosev and Siegmund, 2019; Istook

and Hwang, 2001; Nayak et al., 2015). They serve

as the basis for the design and construction of a 3D

scanner for this application area. The following sec-

tion describes in detail the process of planning and

constructing the scanner in terms of hardware and the

developed software.

3 3D SCANNER SYSTEM DESIGN

While a similar hardware setup was found in a previ-

ous work (Iwayama, 2019), we aimed at a design to

signiﬁcantly lower the entry bar to set up and oper-

ate such a scanner. This is achieved by reducing the

complexity, mainly through the use of standards and

adherence to best practices that work in the environ-

ments of our user groups. We present and discuss the

design of the 3D full-body scanner from three per-

spectives:

• Network, i.e., we describe the setup of the scan-

ner’s hardware components

• Structure, i.e., we report about the scanner’s soft-

ware components and protocols

• Behaviour, i.e., we illustrate how the cameras

work together to take synchronised pictures from

all angles around the subject

Finally, we brieﬂy portray some relevant operational

aspects to complement the design description.

3.1 Network Setup

Our main objective was to develop an easy setup that

adheres to the best practices that work in the envi-

ronments of our user groups. Figure 2 illustrates the

network setup. All Raspberry Pi computers connect

scanner network

uplink network

Internet

Camera

«Raspberry Pi»

camnode

«Raspberry Pi»

centralnode

switch

router

firewall DHCP server Smartphone

GitHub.com Developer PC

User

Developer

0..*

1 0..*

VPN connection

Figure 2: Scanner network setup.

to a switch and form a scanner network, which is con-

nected to an uplink network via a router. The latter

provides network services such as IP address assign-

ment (via the DHCP server) to all Raspberry Pi com-

puters. The user connects the smartphone with the

uplink network, typically via WiFi, to access and con-

trol the scanner. A ﬁrewall shields the uplink network

from the Internet and restricts access. The Internet

hosts a code repository for software provision admin-

istered by a developer.

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

730

The uplink network is an infrastructure network

with an Internet connection through a router. It is

already available in the user’s environment. Typical

uplink networks are in-house university networks or

even everybody’s home networks. In the latter, stan-

dard home routers given to the user by their ISP pro-

vide all the functionality of the uplink network from

our design. As a result, the only setup activity left

to the user is to connect all Raspberry Pi computers

to the switch and then plug it into the router of the

uplink network.

From a network perspective, the scanner does not

form a separate subnet but only consists of several

networked camera devices operating in the same IP

subnet as the uplink network. Therefore, from this

perspective, camera devices do not differ from the

user’s smartphone. In the same way, a user connects

a smartphone to the uplink network, and the user con-

nects the scanner’s camera devices. Consequently, ad-

ditional network management activities such as IP as-

signment and routing are not required.

This is different from previous works, such

as (Iwayama, 2019), where the camera devices are

organised in a separate IP subnet to form a logical

structure, which is then controlled by software. This

adds to the complexity we want to avoid. In the fol-

lowing section, we describe our approach to creating

a logical structure for networked devices.

3.2 Logical Structure

We must know all the networked cameras of the scan-

ner to control them in an organised manner. Thus,

we deﬁne a logical structure. First, we abstract the

concrete hardware by naming the Raspberry Pi com-

puters according to their functional roles. Thereby,

we distinguish between two roles: camnode and cen-

tralnode. The camnode is a Raspberry Pi connected

to a camera. The centralnode is a Raspberry Pi that

controls all camnodes and provides a web-based GUI

to the user via a web server. Figure 3 shows the

UML diagram of the structure. Both nodes con-

nect and exchange messages via the MQTT publish-

subscribe protocol (OASIS, 2019), a well-known de-

facto standard from the Internet of Things (IoT) do-

main. Central to this concept is a broker that col-

lects all messages and forwards them from publish-

ers to subscribers. The MQTT broker is part of the

centralnode. Each camnode connects to the broker

via an MQTT client. The number of camnodes is not

ﬁxed. Zero to many camnodes could be connected to

the broker.

The MQTT broker utilises hierarchically organ-

ised message topics. The ﬁgure 4 shows an ex-

«Raspberry Pi»

centralnode

«Raspberry Pi»

camnode

webserver

MQTT broker MQTT client

GUI

camera

0..*

Figure 3: UML diagram of the abstract model for the scan-

ner’s system structure.

ample of a tree-like topic hierarchy. In this ex-

scanner/

scanner/button

scanner/camnode1/button

scanner/camnode1/image

scanner/camnode2/button

scanner/camnode2/image

...

Figure 4: Example of a MQTT topic hierarchy.

ample, the scanner topic serves as the root and

subsumes the camnodes as subtopics, for example,

scanner/camnode1. Below each camnode topic are

other topics, such as pushing the camera button or

providing the captured image. The scanner/button

topic represents the function of pushing all the camera

buttons simultaneously.

Topics can have two purposes: to publish a mes-

sage to a topic and to subscribe to a topic. The ﬁrst

purpose corresponds to sending a message, and the

second is to receive a message. All message exchange

interactions run through the broker, maintaining the

current state of all the nodes. This tree-like struc-

ture serves in two ways: ﬁrst, as an organisational

structure for collecting and naming all known scanner

hardware components, and second, as an interface to

communicate with the named components.

This approach enables loose coupling. It works

like a marketplace - whoever is on the place, that is,

connected to the broker, is part of the scanner and

provides services, for example, pushing a button or

providing an image. If a camnode fails, the scanner

still works as expected but takes one image less. This

makes the scanner resilient to camnode failure. If a

user installs more camnodes, the scanner seamlessly

integrates their services. No separate software up-

dates are required. This renders the scanner a web

of many interacting cameras, and the structural sys-

tem model enables them to behave like a single large

device.

Evaluation of Low-Cost 3D Scanner Hardware for Clothing Industry

731

3.3 Software Behavior

Using the above structure, the scanner’s behaviour is

determined by a sequence of message exchanges via

MQTT. The UML sequence diagram in Figure 5 il-

lustrates this, which shows the interaction between

a centralnode and all camnodes when capturing im-

ages. Initially, all camnodes subscribe to the topic

User

GUI

«centralnode»

MQTT broker

«centralnode»

MQTT broker

«camnode»

MQTT client

«camnode»

MQTT client

camera

scanner/button

hit bu tto n

publish

but t on to

scanner/button

but t on

take picture

png file

publish png to

scanner/camnode/image

store png

on filesystem

png file on

centralnode

download

png file

get png file

from filesystem

png file

Figure 5: UML sequence diagram for taking pictures.

scanner/button on the central node’s MQTT bro-

ker. In the web browser, the user hits the button on the

web-based UI to take images from all camnodes. This

command is published to the topic scanner/button.

Because all camnodes are subscribed to this topic,

they receive the button command synchronously.

Each camnode then takes an image and publishes it to

the broker under the topic scanner/camnode/image,

where the camnode subtopic is enumerated to distin-

guish the images from each other. The broker re-

ceives all images and stores them as ﬁles in its ﬁle

system. Finally, the user accesses all images via a

web browser for download.

3.4 Operation

We brieﬂy portray some relevant operational aspects

to complement the design description.

3.4.1 Jitter

An important characteristic of a functional 3D scan-

ner is the jitter in the image creation time. The larger

the jitter, the longer the subject must stand still to

avoid blurring of the reconstructed 3D image. All

Raspberry Pi computers have synchronised system

clocks. This allows us to compare the image creation

times of all camnodes. We measured the jitter, that is,

the range across all creation times, in several runs and

found it always below 100ms.

3.4.2 Remote Maintenance and Update

Although the loosely coupled approach of the scan-

ner design makes it resilient and extendable, assis-

tance for maintenance or failure analysis by an ex-

ternal software expert may be required. In Figure 2,

a software developer can access the scanner’s Rasp-

berry Pi computers remotely from the Internet via

VPN and support these activities.

An important part of these activities is updating

the software on the Raspberry Pi computers. The code

was versioned in the Github repository, as shown in

Figure 2. When a code update is requested, the Rasp-

berry Pi reboots and updates its software from the

repository when booting. This allows all Raspberry

Pi computers to be updated simultaneously. Finally,

after the update, the camnode device must reconnect

to the central node MQTT broker to re-establish the

logical structure.

The code repository serves as a single source of

truth. This allows the system to automatically check

that all Raspberry Pi computers work with the same

and the most recent software version. This prohibits

intermittent behaviour in such a distributed system

design caused by deviating software versions.

3.4.3 Single Point of Failure

Both centralnode and camnode are Raspberry Pi com-

puters. The nodes’ roles are not mutually exclusive,

i.e., a Raspberry Pi may work in both roles, centraln-

ode and camnode, at the same time. However, if there

is only one Raspberry Pi with a centralnode role, it

represents a single point of failure. When this device

fails, the MQTT broker disappears and the scanner’s

logical structure dissolves. However, an MQTT bro-

ker can work in the failover mode, that is, if one bro-

ker fails, another redundant one on another Raspberry

Pi takes over. This mitigates the problem of a single

point of failure and does not require additional hard-

ware or software.

3.4.4 Security Considerations

The Raspberry Pi computers are part of the uplink net-

work and are therefore accessible to all devices in this

network, as well as from the Internet. Therefore, we

applied the following measures to secure them with-

out limiting the functionality of the scanner to the

user.

It is recommended and already often seen imple-

mented practice to shield the uplink network from the

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

732

Internet using a ﬁrewall and restrict remote access us-

ing VPN. Even in home networks, ﬁrewalls are in-

stalled on the router provided by the user’s ISP. More-

over, all password-based logins to the Raspberry Pi

computers are disabled, and only cryptographic key-

based logins are allowed. This secures the scanner

from unintended and malicious access from the up-

link network and the Internet. However, physical ac-

cess to these devices may undermine these security

measures.

4 3D IMAGE RECONSTRUCTION

Once the images were captured and downloaded from

3D scanner’s centralnode, post-processing starts to re-

construct the 3D image.

4.1 Image Preprocessing

The preprocessing involves adjusting the images’

brightness, contrast, and colour temperature. This is

crucial to ensure that the images are neither too bright

nor too dark, have sufﬁcient contrast to distinguish

between different objects and have a correct colour

temperature that accurately represents the colours in

the scene.

The second step involves adjusting the exposure of

the images. This is important to ensure that all parts

of the image are correctly exposed and that no areas

are overexposed or underexposed.

The ﬁnal step in the preprocessing pipeline is

white balancing. This is necessary to ensure that

the colours in the images are represented accurately.

White balancing corrects the colours by removing any

colour casts caused by the lighting conditions under

which the images were taken.

For our experiments and to test our hardware and

software architecture, we used in this attempt the free

software Darktable, see ﬁgure 6 to apply these adjust-

ments to multiple images. The preprocessing of im-

ages is a critical step in the photogrammetry work-

ﬂow. By adjusting the brightness, contrast, colour

temperature, and exposure and performing white bal-

ancing, we can signiﬁcantly improve the quality of the

input images and, consequently, the accuracy of the

photogrammetric analysis. Darktable provides a com-

prehensive and user-friendly interface for performing

these preprocessing steps.

4.2 Photogrammetry

This section outlines the post-processing steps and

practical application of RealityCapture (Capturing

Figure 6: Preprocessing in Darktable.

Reality, 2023) in photogrammetry. The process in-

volves the collection of 48 images, alignment of these

images using AprilTag markers, model calculation,

texture application and mesh colourisation. For this

procedure, MATHLAB can also be considered, there

are different solutions for stitching the images (Math-

Works, ided). For this speciﬁc project, a total of

48 images were collected. The quality of these im-

ages signiﬁcantly impacts the accuracy of the ﬁnal 3D

model, so it is essential to ensure that the images are

high-resolution and cover the object of interest from

multiple angles.

The next step is to align the images. This is done

using AprilTag markers, a type of ﬁducial marker sys-

tem. These markers are placed in the scene before im-

age collection, providing a common reference point

across multiple images. For this work, 16 AprilTag

markers were used to align the 48 images accurately.

After the images have been aligned, the next step

is calculating the 3D model. This involves calibrating

the cameras and markers in 3D space. The calculation

uses the aligned images and the known positions of

the AprilTag markers to calculate the 3D positions of

the cameras and to generate a 3D model of the scene.

The ﬁnal step in the process is applying texture

to the 3D model and colourising the mesh. We use

the colour information from the original images to

generate a realistic texture for the 3D model and to

colourise the mesh. This results in a ﬁnal 3D model

that accurately represents the colours and textures of

the original scene.

5 EVALUATION

For the investigation of the generated 3D scans and

their suitability for textile development, the scans of

3 different test persons were ﬁrst visually examined

since artefacts and rough inaccuracies were already

visible at ﬁrst glance.

As seen in Figure 7, especially in the area of the

legs, the inaccuracies are the largest, so the calves are

barely captured, and there are also erroneous connec-

tions between the legs.

Evaluation of Low-Cost 3D Scanner Hardware for Clothing Industry

733

Figure 7: 3D Scans Evaluation.

The inaccuracies on the torso, on the other hand,

are moderate. Still, when the scan is observed visually

alone, it can already be seen that the accuracy and the

mesh do not meet the previously deﬁned requirements

for 3D scans to be used for textile applications. There-

fore, the generated 3D scans are not yet sufﬁcient for

measuring the body dimensions for comparison with

the actual dimensions of the test persons.

6 RESULTS

In this section, we delve into an analytical dis-

course concerning the outcomes derived from the

photogrammetric processes of our 3D scanner. Given

that these constitute our inaugural 3D scans, the re-

sults are beneﬁcial and exhibit a promising trajectory

for future endeavours. For data security and privacy

protection, we truncated the head from the data set.

However, it is noteworthy to mention that, as

pointed out in the evaluation, minor complications

were encountered in the representation of the legs.

Potential resolutions for these issues could include

incorporating additional AprilTag markers, which

would provide more reference points for the align-

ment of images. Alternatively, an enhancement in the

camera resolution could also contribute to rectifying

these issues by capturing more detailed and higher-

quality images.

Looking at the mesh of the scans, we see a very

high amount of data. As shown in Figure 8, the mesh

of test person 1 consists of 48.3K vertices, and the

distance between them is about 0.5 cm.

7 CONCLUSION

In conclusion, this paper presents the design, opera-

tion, and validation of a low-cost 3D scanner system

that utilises Raspberry Pi computers and MQTT pro-

tocol for image capture and processing. The scan-

ner’s software behaviour is determined by a sequence

of message exchanges via MQTT, and the system is

Figure 8: 3D Scan Mesh Vertex distance.

designed to be resilient and extendable, with provi-

sions for remote maintenance and updates. The sys-

tem’s operation is characterised by low jitter in image

creation time, and the potential for a single point of

failure is mitigated by the failover mode of the MQTT

broker. The paper also discusses the post-processing

steps involved in 3D image reconstruction, including

image preprocessing and photogrammetry.

The paper provides a consideration of the scan-

ner’s accuracy and precision and discusses the results

derived from the photogrammetric processes. While

the results are promising, minor complications were

encountered in the representation of the legs, which

could potentially be resolved by incorporating addi-

tional AprilTag markers or enhancing the camera res-

olution.

The previously deﬁned requirements for the scan-

ner used in the textile industry, such as speed and

compatibility, were already met in the ﬁrst attempt.

The criteria that the scanner should capture body di-

mensions and shapes were also met, but not accu-

rately enough for the intended use. The artefacts and

inaccuracies already detected visually are too signiﬁ-

cant for the scans to be used for the generation of mea-

surement tables or for the processing of patterns, but

by optimising the image quality to improve the scan

quality, the next step is to recheck the accuracy of the

scans for measurement and dimension accuracy.

In future work, the system will be further opti-

mised and expanded to improve its performance and

versatility. The scanner represents a valuable tool for

3D image capture and reconstruction, with potential

applications in a wide range of ﬁelds.

REFERENCES

Akca, D., Gruen, A., Breuckmann, B., and Lahanier, C.

(2007). High deﬁnition 3d-scanning of arts objects

and paintings. Optical 3-D measurement technqiues

VIII, 2:50–58.

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

734

Ashdown, S. P., Loker, S., Schoenfelder, K., and Lyman-

Clarke, L. (2004). Using 3d scans for ﬁt analysis.

Journal of Textile and Apparel, Technology and Man-

agement, 4(1):1–12.

Bartol, K., Bojani

c, D., Petkovi

c, T., and Pribani

c, T.

(2021). A review of body measurement using 3d scan-

ning. Ieee Access, 9:67281–67301.

Brake, E., Kyosev, Y., and Rose, K. (2022). 3d garment ﬁt

on solid and soft digital avatars–preliminary results.

Communications in development and assembling of

textile products, 3(2):97–103.

Capturing Reality (2023). Realitycapture. https://www.

capturingreality.com/. Accessed: 2023-07-01.

Chi, L. and Kennon, R. (2006). Body scanning of dynamic

posture. International Journal of Clothing Science

and Technology, 18(3):166–178.

Daanen, H. A. and Psikuta, A. (2018). 3d body scanning.

In Automation in garment manufacturing, pages 237–

252. Elsevier.

D’Apuzzo, N. (2007). 3d body scanning technology for

fashion and apparel industry. In Videometrics IX, vol-

ume 6491, pages 203–214. SPIE.

Franc¸a, J. G. D., Gazziro, M. A., Ide, A. N., and Saito, J. H.

(2005). A 3d scanning system based on laser trian-

gulation and variable ﬁeld of view. In IEEE Inter-

national Conference on Image Processing 2005, vol-

ume 1, pages I–425. IEEE.

Garsthagen, R. (2014). An open source, low-cost, multi

camera full-body 3d scanner. In Proc. of 5th Int. Conf.

on 3D Body Scanning Technologies, pages 174–183.

Haleem, A. and Javaid, M. (2019). 3d scanning applications

in medical ﬁeld: a literature-based review. Clinical

Epidemiology and Global Health, 7(2):199–210.

Hitomi, E. E., Silva, J. V., and Ruppert, G. C. (2015). 3d

scanning using rgbd imaging devices: A survey. De-

velopments in Medical Image Processing and Compu-

tational Vision, pages 379–395.

Istook, C. L. and Hwang, S.-J. (2001). 3d body scanning

systems with application to the apparel industry. Jour-

nal of Fashion Marketing and Management: An Inter-

national Journal, 5(2):120–132.

Iwayama, Y. (2019). Real avatar production-raspberry pi

zero w based low-cost full body 3d scan system kit for

vrm format. In Proceedings of the 10th International

Conference and Exhibition on 3D Body Scanning and

Processing Technologies, Lugano, Switzerland, pages

22–23.

Jhanji, Y. (2018). Computer-aided design—garment de-

signing and patternmaking. In Automation in garment

manufacturing, pages 253–290. Elsevier.

Journalists.org (2020). A guide to photogrammetry photog-

raphy. Accessed: 2023-06-12.

Kyosev, Y. and Siegmund, J. (2019). Evaluation of the accu-

racy and suitability of low-cost rgb-d sensors for auto-

mated air gap measuring in the apparel industry. Pro-

ceedings of 3DBODY.TECH 2019.

MathWorks (Year not provided). triangulatemultiview func-

tion - matlab documentation.

Nayak, R., Padhye, R., Wang, L., Chatterjee, K., and Gupta,

S. (2015). The role of mass customisation in the ap-

parel industry. International Journal of Fashion De-

sign, Technology and Education, 8(2):162–172.

OASIS (2019). MQTT Version 5.0. https://mqtt.org/

mqtt-speciﬁcation/. Accessed: June 27, 2023.

Rocchini, C., Cignoni, P., Montani, C., Pingi, P., and

Scopigno, R. (2001). A low cost 3d scanner based

on structured light. In computer graphics forum, vol-

ume 20, pages 299–308. Wiley Online Library.

Rozmus, M., Tokarczyk, J., Michalak, D., Dudek, M.,

Szewerda, K., Rotkegel, M., Lamot, A., and Ro

ser,

J. (2021). Application of 3d scanning, computer sim-

ulations and virtual reality in the redesigning process

of selected areas of underground transportation routes

in coal mining industry. Energies, 14(9):2589.

RPi (2023a). Raspberry Pi 4 Model B by the Raspberry Pi

Foundation. https://www.raspberrypi.com/products/

raspberry-pi-4-model-b/. Accessed: June 27, 2023.

RPi (2023b). Raspberry Pi NoIR Camera V2 by the Rasp-

berry Pi Foundation. https://www.raspberrypi.com/

products/pi-noir-camera-v2/. Accessed: June 27,

2023.

Schenk, T. (2005). Introduction to photogrammetry. The

Ohio State University, Columbus, 106:2005.

Schlich, E., Schumm, M., and Schlich, M. (2010). 3d-body-

scan als anthropometrisches verfahren zur bestim-

mung der speziﬁschen k

orperoberﬂ

ache. Ern

ahrungs

Umschau, 57(4):178–183.

Sepasgozar, S. M., Shirowzhan, S., et al. (2016). Challenges

and opportunities for implementation of laser scanners

in building construction. In ISARC. Proceedings of the

international symposium on automation and robotics

in construction, volume 33, page 1. IAARC Publica-

tions.

Spelic, I. (2020). The current status on 3d scanning and

cad/cam applications in textile research. Interna-

tional Journal of Clothing Science and Technology,

32(6):891–907.

Vedantu (2023). Photogrammetry. Accessed: 2023-06-12.

Yahav, G., Iddan, G. J., and Mandelboum, D. (2007). 3d

imaging camera for gaming application. In 2007 Di-

gest of Technical Papers International Conference on

Consumer Electronics, pages 1–2. IEEE.

Zeraatkar, M. and Khalili, K. (2020). A fast and low-cost

human body 3d scanner using 100 cameras. Journal

of Imaging, 6(4):21.

Zyła, K., Kesik, J., Santos, F., and House, G. (2021). Scan-

ning of historical clothes using 3d scanners: Compar-

ison of goals, tools, and methods. Applied Sciences,

11(12):5588.

Evaluation of Low-Cost 3D Scanner Hardware for Clothing Industry

735