Artificial intelligence technologies for more flexible recommendation in uniforms
| DOI | https://doi.org/10.1108/DTA-09-2021-0230 |
| Published date | 04 January 2022 |
| Date | 04 January 2022 |
| Pages | 626-643 |
| Subject Matter | Library & information science,Librarianship/library management,Library technology,Information behaviour & retrieval,Metadata,Information & knowledge management,Information & communications technology,Internet |
| Author | Chih-Hao Wen,Chih-Chan Cheng,Yuh-Chuan Shih |
Artificial intelligence technologies
for more flexible recommendation
in uniforms
Chih-Hao Wen
Department of Communications Management, Shih Hsin University,
Taipei, Taiwan, and
Chih-Chan Cheng and Yuh-Chuan Shih
Department of Logistics Management, National Defense University, Taipei, Taiwan
Abstract
Purpose –This research aims to collect human body variables via 2D images captured by digital cameras.
Based on those human variables, the forecast and recommendation of the Digital Camouflage Uniforms (DCU)
for Taiwan’s military personnel are made.
Design/methodology/approach –A total of 375 subjects are recruited (male: 253; female: 122). In this study,
OpenPose converts the photographed 2D images into four body variables, which are compared with those of a
tape measure and 3D scanning simultaneously. Then, the recommendation model of the DCU is built by the
decision tree. Meanwhile, the Euclidean distance of each size of the DCU in the manufacturing specification is
calculated as the best three recommendations.
Findings –The recommended size established by the decision tree is only 0.62 and 0.63. However, for the
recommendation result of the best three options, the DCU Fitting Score can be as high as 0.8 or more. The
results of OpenPose and 3D scanning have the highest correlation coefficient even though the method of
measuring body size is different. This result confirms that OpenPose has significant measurement validity.
That is, inexpensive equipment can be used to obtain reasonable results.
Originality/value–In general, the method proposedin this study is suitablefor applicationsin e-commerceand
the apparelindustry in a long-distance,non-contact andnon-pre-labeled mannerwhen the world is facing Covid-
19. In particular,it can reduce the measurement troublesof ordinary users when purchasing clothing online.
Keywords Body measurements, Body detection, Decision tree, Fit customization, Anthropometry
Paper type Research paper
1. Introduction
When soldiers wear unfit uniforms, they may fall into machinery or accidents because of the
too loose size of uniforms, or their movement may be restricted owing to the too-tight size.
Hence, the correct size of uniforms is vital for soldiers (Kolose et al., 2021a). The body shape
variables, including girth, length and width, can be obtained through direct or indirect
measurement methods, in which the appropriate size of military uniforms is recommended.
Body measurements to measure clothing are generally obtained manually using a
measuring tape with the assistance of a second person (Senanayake et al., 2018). This
measurement method is very time-consuming and labor-intensive. It will also affect the
measurement results due to the proficiency of the measurement personnel, which causes
errors (Lin and Wang, 2011). In addition, these restrictions might result in inaccurate artificial
sizes, then leading to unfit clothing (Liu et al., 2014).
Recently, the application of indirect measurement and collection of human body
measurement data through various 3D scanning instruments has become very common (Wen
and Shih, 2021). As a result, the collection, evaluation and update of human measurement
data are also completely changed (Kolose et al., 2021b). The available 3D body scanners are
DTA
56,4
626
Funding: No funding has been received for this research.
Declaration of Interest Statement: Authors declare no conflict of interest.
The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/2514-9288.htm
Received 1 September 2021
Revised 1 December 2021
Accepted 17 December 2021
Data Technologies and
Applications
Vol. 56 No. 4, 2022
pp. 626-643
© Emerald Publishing Limited
2514-9288
DOI 10.1108/DTA-09-2021-0230
light-based, laser-based or scanners that use motion sensors (such as Microsoft Kinnect®)to
take body measurements. The data obtained by this measurement method is more effective
and accurate than traditional measurement methods (Lacko et al., 2017). Many related studies
have obtained data through 3D scan and used it as the basis for the design of bulletproof vests
(Wen and Shih, 2021), clothing (Markiewicz et al., 2017), gloves (Yu et al., 2021), etc. It is further
used as an application for predicting clothing size for recommendations (Daanen, 2014;
Kolose et al., 2021a,b). However, the expensive cost of indirect measurement equipment, the
uneasy equipment maintenance, the complex parameter adjustment and the weak mobile
capabilities are the current concerns (Senanayake et al., 2018). Besides, the operation and data
processing of the 3D scanning system require special software to match, which may also
affect the accuracy of its data capture.
Now, many studies have begun to use cheap digital cameras to obtain data of 2D images
efficiently. These data use the silhouette method or map each pixel to the hue saturation
value(HSV)modeltoextractthefeaturesofthebody(Lin and Wang, 2011;Senanayake
et al., 2018;Seo et al.,2006;Shah et al.,2019), or carry out clothing-related applications
(Meunier and Yin, 2000;Sun et al.,2017).In the data acquisition process, it is necessary to set
the background of the experimental environment or use more than one lens. For real-world
applications with complex backgrounds or e-commerce applications, there may be more
interference variables.
Using artificial intelligence to predict body dimensions rather than measuring them
physically is a new research direction in the apparel industry (Liu et al., 2017). OpenPose in
artificial intelligence does not require background settings or pre-labeling. Instead, after
capturing images through cameras, mobile phones, webcams, etc., it can directly identify the
position of human bones (Kim et al., 2021;Xu et al., 2020), actions (Domingo et al., 2021;Nose
et al., 2019;Vasconez et al., 2021) and hand gestures (Mazhar et al., 2019). Related research also
compares OpenPose with other sophisticated measurement equipment (Clark et al., 2019;Kim
et al., 2021;Ota et al., 2020,2021), in which, the results show that OpenPose’s 2D analysis can
provide valid values for humans with frontal and non-moving postures.
The Digital CamouflageUniforms (DCU) used by Taiwan’smilitary(showninPlate 1)isa
two-sexes cloth ing system. The male D CU’sshirt codes are odd numbers rangingfrom 35 to 61
(14 types), and females are even numbers ranging from 32 to 56 (13 types). All codes are
provided in S, M and L sizes. The number of the DCU roughly represents the chest
circumference(inches); the length is adjustedby S, M and L. Therefore, when recruitsenter the
army, the petty officer estimates the recruits’initial shirt size by their chest circumferences.
Then, different clothing sizes are provided to try on until the right DCU is got. The average
processingtime per person is about 10 min. However, thenumber of reported recruits is about
200,and the total time for trying on clothesis limited to 2 h. Especially, it is challengingto adjust
the recruits’clothing when it iscarried out with other activities(such as physical examination)
simultaneously.OpenPose’s automatic measurement derived from bodyscanning technology
may improve the efficiency of the DCU adaptationfor recruits in Taiwan’sarmy.
The research motivation is summarized as follows:
(1) The activities of military personnel are affected by the DCU, so a suitable size DCU is
essential for the military personnel.
(2) For expensive body-shaped data by the 3D measurement method, the incapability of
mobility on the 3D measurement equipment, and errors caused by different manual
measurements and the measurement personnel experience, the benefits of the DCU’s
recommendations are affected. Therefore, it is necessary to have a low-cost and low
error method for obtaining body shape data in practice.
Computer
vision for
recommended
uniforms
627
To continue reading
Request your trial