�ݺ�ߣ

Fast, portable & low-cost
3D foot digitizers:
validity & reliability
Alfredo Ballester

IBV is a private not-for-profit R&D organisation
Ergonomic design consultancy
for manufacturing industries
Research & Development partner
for IT companies and Academia
Apparel Sports Transport
Health
Safety
Leisure
Appliances Elderly
Orthotics
Data-driven 3D modelling
from scans (3D) images (2D)
or body measurements (1D)
Data analytics & services
measuring, shape analyses, size advice
Applications & hardware prototyping

Introduction
3D Foot solutions
3D Body solutions
Conclusions

8th 3D BODY.TECH Conference & Expo
Montréal, Canada, 11-12 October 2017
Fast, portable and low-cost 3D foot digitizers:
validity and reliability of measurements
Why should we care about reliability?
Size Selection
Fit prediction
Virtual Simulation
Made to Measure
Design of sizing systems
…
ingredients for
effective
ergonomic design
- modelling body to product
interaction
- reliable product information
- reliable body information

3D scanning of living bodies
3D scans are STATIC & RIGID

3D scans are STATIC & RIGID Living bodies are SOFT & ARTICULATED

Living bodies are SOFT & ARTICULATED3D scans are STATIC & RIGID
≠ pose
≠ muscle
contraction
≠ respiration
≠ garments

Introduction
3D Foot solutions
3D Body solutions
Conclusions

3D foot measuring solutions
Conceived for retail and
serial measuring

3D foot measuring solutions
Conceived for retail and
serial measuring
Conceived for home and
hardware free locations

V1.x
(2015)
Premiere 3DBODY.TECH2017

V1.x
(2015)
Premiere 3DBODY.TECH2017
ankle
V2.0
(2017)

Data-driven 3D reconstruction of feet
Image Acquisition

Image Acquisition
Image Segmentation

Image Acquisition
Image Segmentation
Camera Calibration

Image Acquisition
Image Segmentation
3D Reconstruction
3D Foot Model
Camera Calibration
Data-Driven
Space of Shapes of
Human Foot

Image Acquisition
Image Segmentation
3D Reconstruction
Digital Measuring
Foot Measurements
3D Foot Model
Camera Calibration
Data-Driven
Space of Shapes of
Human Foot

Limitations of Data-driven 3D reconstruction
Lack of
separated toes
Ankle (grey) not
actively reconstructed
Better barefoot than socks
Not suitable for feet with
severe deformations or amputations

Design of the Experiment
 16 subjects
• 8 Females & 8 Males
• 24-63 y.o
• Sizes EUR 35-47 (US 5-13)
 4 measuring methods
• Expert manual measuring
• 3D laser scanner (Infoot)
• DomeScan
• 3D avatar app
 3 non-consecutive repeated
measuremetns

Data Analyses
 Reliability of measurements (Eliasziw et al. 1994)
• Single measuring method model
• Intraclass correlation coefficient (ICC)
• Standard Error of Measuremetn (SEM) SEM = 𝜎𝑒 = 𝑀𝑆𝐸 𝐼𝐶𝐶 =
𝜎𝑆
2
𝜎𝑆
2
+ 𝜎𝑒
2
𝑖 = 1, … , 16; 𝑘 = 1, 2, 3
𝑥𝑖𝑘 = 𝜇… + 𝜋𝑖 + 𝜖𝑖𝑘
𝑥𝑖𝑗𝑘 = 𝜇… + 𝜋𝑖 + 𝛾𝑗 + 𝜋𝛾 𝑖𝑗 + 𝜖𝑖𝑗𝑘
𝑖 = 1, … , 16; 𝑗 = 1, … , 4; 𝑘 = 1, 2, 3
 Compatibility of measurements across methods
(Eliasziw et al. 1994)
• Inter measuring method model
• Systematic errors are fixed method effects 𝛾𝑗
 Reliability of 3D surface
• Average surface-to-surface distance between repetitions
 Accuracy of 3D surface compared to laser scanner (Exploratory)
• Average surface-to-surface distance between Dome/app repetitions
and Infoot repetitions

Relative reliability of measurements (ICC)
Relative
reliability %
Expert manual
measurements 3D scanner DomeScan Avatar3D
Expert manual
measurements
(other studies)
3D Scanner
(other studies)
Foot Length FL .998 .999 .999 .999 .968-.998 .997-.998
Toes Girth TG .977 .981 .988 .993 - -
Ball Girth BG .991 .994 .997 .998 .991 .995-.996
Ball Width BW .990 .992 .995 .996 .871-.992 .950-.991
Instep Girth IG .997 .999 .996 .997 .990-.992 0.999
Instep Height IH .977 .985 .988 .985 .897-.994 .970-.990
Toes Height TH - .922 .928 .918 - -
Ball Height BH - .968 .976 .964 - -
Instep Width IW - .991 .995 .992 .978-.992 -
Heel Width HW - .998 .992 .989 .969-.992 .870-996

Absolute reliability of measurements (SEM)
Random error in
millimeters (mm)
Expert manual
measurements 3D Scanner DomeScan 3D Avatar
Expert manual
measurements
(other studies)
3D Scanner
(other studies)
ANSUR max.
allowable error
Foot Length FL 0.9 0.8 0.6 0.7 0.5-2.4 0.8-1.0 3
Toes Girth TG 3.3 3.0 2.3 1.8 - - -
Ball Girth BG 2.1 1.6 1.2 1.0 1.8-2.1 1.1 4
Ball Width BW 0.8 0.8 0.6 0.6 1.0-1.8 0.7 2
Instep Girth IG 1.1 0.6 1.3 1.2 - 0.6 -
Instep Height IH 1.0 0.8 0.7 0.8 0.5-1.8 0.5-0.6 -
Toes Height TH - 0.7 0.7 0.7
- - -
Ball Height BH - 0.7 0.6 0.8
1.0-1.3 0.8-1.1 -
Instep Width IW - 0.8 0.6 0.7
- - -
Heel Width HW - 0.3 0.6 0.7
1.3 0.3-0.4 2DomeScan
and the app
comparable
to Infoot

Compatibility across methods (bias)
Systematic
Error (bias)
in mm
Avatar3D –
3D Scanner
DomeScan –
3D Scanner
Trad.Ant. –
3D Scanner
ISO 20685:2010
Compat. of
measuremetns
Foot Length FL 1.0* -0.2 -0.5* 2
Toes Girth TG 0.2 -1.2 -5.9* 2
Ball Girth BG -0.1 -1.0 2.3* 2
Ball Width BW 0.1 -1.0* -1.2* 2
Instep Girth IG 0.3 -0.7 -1.1* 2
Instep Height IH -0.3 0.3 -2.0* 2
Toes Height TH -0.6* 0.2 - 2
Ball Height BH -0.1 0.7* - 2
Instep Width IW 0.5* -0.8* - 2
Heel Width HW 0.3 0.6 - 2

2 mm
0 mm
1 mm
Infoot
mean = 0.32 mm
RMS = 0.50 mm
DomeScan
mean = 0.40 mm
RMS = 0.55 mm
3Davatar
mean = 0.43 mm
RMS = 0.60 mm
Average reliability of 3D foot surface

Average Accuracy of 3D foot surface (exploratory)
2 mm
0 mm
1 mm
DomeScan
mean = 0.82 mm
RMS = 1.02 mm
3Davatar
mean = 0.81 mm
RMS = 1.02 mm
Frequency of surface to
surface distance

IBV’s 3D Body solutions
3D to 3D
Watertight 3D body models with
anatomical slit modelling from
raw scans from any vendor

2D to 3D
3D body models from two
photos (front and side) taken
with a smartphone app
3D to 3D

2D to 3D
3D body models from two
photos (front and side) taken
with a smartphone app
3D to 3D
Measuring, Rigging & More!
Any of the 3D body models from 2D-3D or 3D-3D can be
measured, rigged with a 25-bone skeleton and posed or
animated. They can be equipped with custom skin
topologies and enriched with geometries and textures

Raw 3D scans to 3D models
Data-driven anatomical slit modelling,
denoising & geometry completion
High resolution Low resolution

Data-driven 3D reconstruction
Images, gravity sensor &
camera parameters
Image Segmentation
3D Reconstruction
Measuring, rigging, etc.
3D Body Model
Camera Calibration
Data-Driven
Space of Shapes of
Human Body
Guiding outline
Data-Driven
Space of Body
outlines
Gender, age, weight, height

Data-driven 3D reconstruction of bodies
2D-3D
model
3D-3D
model
Actual
scan

Actual
scan
2D-3D
model
3D-3D
model

Conclusions
 Reliability of foot measurements and 3D surfaces are
comparable to actual 3D scans
 Compatibility of foot measurements is granted
between reliable methods
 High accuracy of the 3D surface

Conclusions
 Reliability of foot measurements and 3D surfaces are
comparable to actual 3D scans
 Compatibility of foot measurements is granted
between reliable methods
 High accuracy of the 3D surface
 Highly accurate 3D-3D modelling & measuring
 Expected reliability of 2D-3D measurements comparable
to actual body scanners with centimetric precision
 Visually accurate 2D-3D reconstructions
 3D avatars/models ready to interface other applications
3D-3D 2D-3D

Thank you!
Ana Piérola
Sandra Alemany
Eduardo Parrilla
Jordi Uriel
Cristina Pérez
Ana V. Ruescas
Vicent Ortiz
Beatriz Nácher
Juan C. González
Álvaro F. Page
Mateo Izquiero
…
Alfredo Ballester
alfredo.ballester@ibv.upv.es

�ݺ�ߣ

Fast, portable & low cost 3D foot digitizers: validity & reliability @3DBody.Tech 12th Oct 2017

More Related Content

Fast, portable & low cost 3D foot digitizers: validity & reliability @3DBody.Tech 12th Oct 2017