Journal of Physics: Conference Series

PAPER • OPEN ACCESS You may also like

- The U.S. consumer phosphorus footprint:
Digital image processing techniques applied to where do nitrogen and phosphorus
diverge?
pressure analysis and morphological features Geneviève S Metson, Graham K
MacDonald, Allison M Leach et al.

extraction in footprints. - Comparison of food supply system in

China and Japan based on food nitrogen
footprints estimated by a top-down method
To cite this article: F J Buchelly et al 2016 J. Phys.: Conf. Ser. 705 012020 Junko Shindo, Azusa Oita, Kentaro
Hayashi et al.

- Assessment of equilibrium field coil

misalignments on the divertor footprints in
NSTX-U
View the article online for updates and enhancements. S. Munaretto, T.E. Evans, N.M. Ferraro et
al.

This content was downloaded from IP address 190.96.98.240 on 23/12/2022 at 22:03

SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

Digital image processing techniques applied to pressure

analysis and morphological features extraction in footprints.

F J Buchelly1, 2, 3, D Mayorca3, V Ballarín1 and J Pastore1

1
Laboratorio de Procesamiento Digital de imágenes, ICyTE (Instituto de
Investigaciones Científicas y Tecnológicas en Electrónica) Facultad de Ingeniería,
Universidad Nacional de Mar del Plata, Argentina
2
Consejo Nacional de Investigaciones Científicas y Técnicas - CONICET
3
Grupo de Óptica y Láser, Departamento de Física, Universidad del Cauca, Colombia

E-mail: [email protected]

Abstract. This paper shows the correlation between foot morphology and pressure distribution
on footplant by means of a morphological parameters analysis and pressure calculation.
Footprint images were acquired using an optical pedobarograph and then processed for
obtaining binary masks and intensity images in gray scale. Morphological descriptors were
obtained from the binary images and the Hernandez Corvo (HC) index was automatically
calculated for determine the type of foot. Pressure distributions were obtained from gray scale
images making a correspondence between light intensity in footprints and pressure. Pressure
analysis was performed by finding the maximum pressure, the mean pressure and the ratio
between them that determines the uniformity of the distribution. Finally, a high correlation was
found between this ratio and the type of foot determined by HC index.

1. Introduction
Obtaining footprint features is important issue in health and sports, because it permits to study
pathologies related to foot morphology [1] [2], human locomotion [2] [4], or even in detecting risks
produced by diabetes [5]. These pathology diagnostics are decisive in many disciplines like preventive
medicine [6], clinical medicine [7], orthopedics [8], sports medicine [9] [10], and other.
Morphological and pressure features are data of medical interest. Currently, morphology features are
visually obtained on footprint images in manual way [10], and also in semiautomatic way [11]. There
are different forms of obtaining a person’s footprint. Lara et al. [9] show some of them. Commonly,
ink based methods are used, in which skin is put in contact with a paper sheet or some film. The
patients not only feel uncomfortable with this procedure, but it is possible to make mistakes obtaining
footprint images because diagnoses are subjective and it´s difficult to estimate pressure values.
A different system is a parallelepiped footprint viewer, or podoscope, with a flat glass surface in which
a person keeps a vertical position supporting both feet, like is used in Oller [3]. In that work,
fluorescent lamps emit yellow light at both sides of the glass, showing a luminous footprint when the
total internal reflection condition (TIR condition) is eliminated by contact of the skin and glass. A
similar version, proposed by Buchelly et al. [2] and Mayorca [5] uses green light, instead of yellow,
that improves footprint intensity, due to radiation absorption of the glass that minimizes its value for
near green wavelengths, allowing simpler Digital Image Processing.

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

Pressure distribution measurement can be made by sensor arrays [12] [13]. Usually, their spatial
resolution is poor and can lose information about foot form in places where pressure value is low.
Some other works [2] study feet morphology and pressure using different devices. Bates et al. [14]
shows that geometry of a footprint, done in a compliant material, is related to pressure distribution.
According to the previous conclusion, using a podoscope like the ones described in [2] and [5] also
allows to determine the pressure of each point in a foot plant considering that skin roughness is
deformed depending on the supported pressure. This deformation influences strongly the light
intensity and allows integrate both devices in one called “pedobarograph”.
Following the previous ideas, in this work we propose a comparative analysis of Digital Image
Processing techniques used to automatically determining morphological parameters associated to each
type of foot and pressure distribution in footprint images acquired from a pedobarograph device,
respect to similar previous works.

2. Foot pathologies
From footprint data it is possible to make the following pathological studies.

2.1. Footprint morphology

There are several classifications of feet morphologies. One of them is the rank from plane foot to
extreme cavus foot. Figure 1 shows examples of this rank. Each one produces a particular footprint
and pressure distribution, affecting the gait and making feel more or less pain when the person stands
in two feet. An intermediate interval between these extreme states is considered as normal foot.

Figure 1. Example of footprints of plane foot (up) in different grades,

normal foot (medium left) and cavus foot (down). Image taken from [15]

2.1.1. Plane foot

A plane foot occurs when the decreasing of the plant arc is below normal values according to the
Costa-Bartani angle and the location of scaphoid bone respect to the Feiss line. The ball is configured
depending on these parameters, modifying the support points and the plant morphology [16].

2.1.2. Cavus foot

The cavus foot is often characterized by a ball accentuation due to the height of the longitudinal arcs.
The distribution of pressure on the floor results perturbed by the oncoming of the anterior and
posterior supports as well as the reduction of the support surface, mainly at the expense of the lateral
border, which serves of basis for podoscopic classification of cavus feet [17].

2.1.3. Determining type of foot

Some authors [5] [9] [10] [11] [18] use some indexes that permit assign a value to each type of foot by
its respective footprint. Some of them are: Clarke’s angle, Chippaux`s angle, Arc Index, and others.
However, the most used is the Hernandez Corvo (HC) index, also mentioned in the previous
references. In order to assess the HC index, the following steps are performed [9]

2
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

Figure 2. Hernandez Corvo procedure for footprint

assessment.

 Mark two points in the inner salience (1 and 1’) and do the “initial trace”, joining both points
 Mark a point in the most anterior part of the footprint (including toes) and other in the most
posterior point (2 and 2’)
 Trace perpendicular lines to 2 and 2’ respect to the initial trace. The distance between this
trace and point 1 is the “fundamental length” and it has to be taken many times as possible on
the initial trace (3, 4 and 5)
 Trace a line perpendicular to 3, crossing by the outer point of the footprint, and other two lines
in similar way perpendicular to 4 and 5 in the outer part (6, 7 and 8, respectively). The
distance between initial trace and 6 is X (metatarsial width); the distance between 9 and 7 is Y
(extern arc, support surface in the middle of foot)
 Type of foot is obtained using Equation 1:

(1)

HC index value classify foot in following percent ranges [10]

Table 1. Type of foot depending on

HCI value.

IHC (%) TYPE OF FOOT

0-34 Plane
35-39 Plane-Normal
40-54 Normal
55-59 Cavus-Normal
60-74 Cavus
75-84 High Cavus
85-100 Extreme Cavus

2.2. Pressure Distribution

The barometric parameters of foot are not independent of the morphology. The simplest example is the
variation of pressure in different types of footprints determined by HC index. In extreme cavus foot,
standing people feel more pain because corporal weight is supported by a smaller area. Another case
in which morphology is related to pressure is Hallux Abductus Valgus (HAV) or “bunion”, where not
only the deformation of foot causes nuisance but also changes pressure distribution [19].
On the other hand, plant pressure analysis allows determining not only the morphologic pathologies,
but also diabetic foot associated with diabetes mellitus. Among risk factors that can become in diabetic
foot, the main pathology is the occurrence of abnormal pressure regions in foot [20] and redistribution

3
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

of maximum pressure points. Those signs and many others may allow the generation of ulcers and
even amputation [21] [22].

3. Methodology
The proposed processing system is summarized in figure 3

Gray Thresh.
Pressure
Scale 1 1 Product
Analysis

Gray Thresh.
Scale 2 2
Image Morphology
Capture Preprocessing AND Analysis
Gray Thresh.
Scale 3 3

Gray Thresh.
Scale 4 4

Figure 3. Flowchart that describes the general steps of the proposed methodology.

3.1. Image acquisition

In this work, a pedobarograph was fabricated to generate visible footprints, similar as the proposed in
works [2] [5] [11] [23] [24] [25]. Its image formation system is described in figure 4.
Some of the mentioned devices use a film between foot skin and glass (or acrylic) with the purpose to
avoid the light noise of the environment and reduce the dependency between the generated luminosity
and the skin humidity. In spite of these advantages, using that material implies other kinds of noise
and disadvantages because of its texture (loss of information), adherence with acrylic or glass,
problems due to its thickness, and other [23]. For that reasons, this work doesn’t use a transducer film,
it makes footprint detection by applying different transformations to grayscale and creating binary
masks by means of thresholding.

Figure 4. Image acquisition system by means of frustrating total internal reflection of light.

3.2. Preprocessing
The main objective in many Digital Image Processing applications is to measure parameters from the
real world. In this case, it is necessary that the geometry of the acquired footprint image corresponds

4
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

to the geometry of flat image in glass for determining morphological features (lengths, areas)
correctly. The camera was calibrated to correct optical distortions (radial, tangential, center deviation)
produced by the camera lens [26] [27]. The camera parameters were obtained and correction model
was generated by using a chessboard pattern. Additionally, it was needed to perform a geometric
transform because of the camera position and using a flat mirror introduces perspective effects.
Sometimes noise filtering is used to avoid false detection and abrupt changes in gray levels. However,
low pass filters mix light from FTIR with ambient light, affecting the footprint detection. Low pass
filters are used later, after footprint is detected, to facilitate the study of pressure distribution. All the
algorithms described below were developed in MatLab®.

3.3. Transformation to grayscale

This step is based on the use of RGB and HSV color spaces, having accounted the meaning of green
color over all other. Mayorca [5] use an opaque transducer film to simplify the detection with a gray
background and a green footprint, allowing only to intensity value and saturation value define the
threshold method (V and S in HSV space). In this work, that method was omitted because of the non
uniformity of background. Four independent transforms were used to highlight the footprint regions,
instead. These transforms are explained in following sections, considering a color pixel as a three
dimensional vector whose values in each component are two dimension functions (color planes) of
spatial coordinates (Equations 2 and 3) and a gray scale pixel is the value of a two dimension function
of spatial coordinates, resulting of any operation with color planes. The range of gray levels are
normalized (from 0 to 1) instead of the usual (0 to 255) range.

IRGB(x,y)=(R(x,y), G(x,y), B(x,y)) (2)

IHSV(x,y)=(H(x,y), S(x,y), V(x,y)) (3)

3.3.1. Subtraction of red and blue values

The first method of green highlighting is the one used in references [2] [11], defined by equation 4.

(4)

3.3.2. Gaussian of Hue values

Hue plane of HSV space is transformed to an intensity distribution according to Gaussian function
with a mean in green ( =1/3), defined by equation 5, and with an adjust factor f(x,y).

(5)

3.3.3. Subtraction of mean RGB from V value

In HSV space, V is defined as the maximum of R, G, and B [28], then there is a high difference
between V and the mean of R, G, and B when one of them is 1 and the other values are 0. In the other
hand, the difference is low if R, G and B are similar. This difference is defined by equation 6.

(6)

3.3.4. Comparison between R, G, B

As it was shown previously, the predominant color in footprint is green, although it can be mixed with
a few of blue or red. Without compare proportions of blue and red present in a footprint, a fourth
grayscale image can be defined under the condition given by equation 7.

(7)

5
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

In equation 7, α and β are adjust constants, fixed for all images. For this work these values are α =1.5
and β =1 and were found experimentally.

3.4. Thresholding and Morphological operations

After gray scale transformations are performed, each image is thresholded with an optimum value
obtained empirically. Although every gray scale image show an intensity pattern, where the brightest
pixels represent green regions in the original image, the results are not the same. The obtained binary
masks usually are different and a definitive mask by means of a logical AND operation was generated.
Thus, both noise and morphological differences are eliminated.
Occasionally, some isolated small regions still remain, or some borders have defects or connected
objects have holes. These aspects were corrected using morphological operations. We applied an
opening followed by closing to smooth borders [29], delete isolated and small undesired particles. We
used a disk structuring element. Finally, an algorithm of holes filling is applied.

3.5. Foot morphology analysis

In this work, the equivalence between physical lengths on the glass surface and discrete lengths in the
images was calculated (1 pixel-length=0.054cm). This conversion parameter is the same for vertical
and horizontal directions in images. With this value, is possible to measure patient’s feet length and
width, as well as orientation of the feet, arc length, support area, although the decisive parameter to
determine the type of foot was Hernandez Corvo Index, automatically calculated.
First step in determining HC index was to align the images of each foot independently along of the
major axes of the feet, in approximation of the principal length direction. This line was in vertical
orientation in each image. Then, X and Y were found like horizontal lines and HCI was determined
using Equation 1. Nevertheless, that procedure sometimes causes errors due to the unsupervised
detection of extern in the foot arc. Because of this, two alternative methods were performed for the
HCI determination:
 Second method: taking the X value as the foot width and the Y value as the width of the
footprint at the medium height between points 2 and 2’.
 Third method: taking the X value as the maximum horizontal width in the anterior part of foot,
taking the X’ value as the maximum horizontal width in the posterior part of foot and the Y
value as the minimum horizontal width between X and X’.
The developed system displays HCI calculation using the three methods and the average value.

3.6. Pressure analysis

A useful intensity distribution Igreen(x,y) is obtained by multiplying the generated binary mask and a
gray scale image. The best way to represent the variation of the green intensity in footprints was Gray1
(See equation 4). The image Gray2 may have high values even if the green intensity is low, because it
is calculated using Hue. Gray3 and Gray4 may have abrupt intensity spatial variations that are not
considered for this purpose.
From this intensity distribution it is possible to determine pressure values based in previous works. In
this work, a proportional allocation was used, as proposed in [2], where the person’s weight (w=m.g,
product between mass and gravity acceleration) is distributed according to each pixel value by means
of equation 8, proposed by Buchelly et al. [2].

(8)

Mayorca [5] shows that pressure versus intensity dependency present high correlation, so the
dependency is linear. In our work, we use equation 8 because a linear dependency can be transformed
to a proportional allocation performing an offset compensation.

6
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

4. Results and Discussion

In this work, the proposed system is evaluated with images of 11 different people footprints including
children, adult men and adult women. The footprint image segmentation of a left foot is presented as
an example in Figure 5, as well as its intensity distribution.

b) c) d) e)

a) j) k)

f) g) h) i)
Figure 5. Footprint segmentation and intensity distribution: a) Part of the original image
corresponding to a left foot, b) Gray1, c) Gray2, d) Gray3, e) Gray4, f) to i) binary images obtained by
thresholding, j) AND operation between the binary images and k) intensity distribution.
The procedure applied for obtaining the final mask was simplified by thresholding only Gray1 image,
like the approach of [11]. However, some undesired green regions are captured (figure 5f) to 5i) near
the heel) that may be large or small. In [11], those blobs were deleted after size and form analysis. In
our work, they were eliminated or attenuated using the AND operation. If undesired regions remained,
they were removed by means of morphological opening and closing.

Figure 6. Comparison of the three methods for determining HC index and the average value. First half
of data are right feet and second half are left feet.

7
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

In order to calculate the HC index, three mentioned methods were used as described before and then
the average was calculated. They present some differences, as shown in Figure 6; however, the
performance of each method is similar to the others, showing a high correlation.
In the same way, the Manual calculation of HC index is performed and the mean deviation from each
method respect to the manual result is calculated, as shown in Table 2.

Table 2. Comparing performances of the three methods for HC index calculation.

Method 1 Method 2 Method 3 Average

Square root of the mean square deviation of each
9.78% 9.78% 14.98% 9.43%
method respect to the manual procedure.

On the other hand, the pressure distribution is found using equation 5. As an example, figure 7 shows
pressure distributions for a plane foot, a normal foot and a cavus foot.

Figure 7. Pressure distribution of a plane foot (left) a normal foot

(center) and a cavus foot (right).

The maximum and average values of pressure in each image varies depending on person’s weight, foot
size and foot type. Figure 8 shows the average pressure and the maximum pressure of each footprint.

Figure 8. Mean and maximum pressure value for each sample foot.

Although the previous values are important for clinical assessments, a better diagnostic parameter may
be the ratio between maximum pressure and mean pressure, because it measures the uniformity of the
pressure distribution. In the analyzed samples, the minimum values of the ratio belong to the right and
the left foot of the same person, and the HC index shows that both are plane. Additionally, this person
is the younger child (samples 8 and 19 of figure 8). On the other hand, the maximum values of this
ratio belong to extreme cavus feet (samples 6 and 17), showing high correlation with the HC index.

5. Conclusions
An automatic digital image processing system was created for detecting footprints from an optical
pedobarograph for analyzing foot plant morphology and pressure distribution. Four transforms to gray
scale were used for segmenting the footprint region with fixed thresholds. Definitive masks were
obtained performing a logical AND operation that eliminates false detected binary objects. The

8
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

remaining undesired regions were removed and borders were smoothed by means of morphological
opening and closing.
Length measures can be done using a conversion factor between physical lengths in centimeters and
image lengths in pixels, as well as area measurements, for footprint morphological analysis in a binary
mask. Hernandez Corvo assessment was performed automatically for twenty two samples (eleven
persons) and classifying each foot as plane, normal or cavus in different grades. Three methods were
used for determining HC index and the average value was calculated. These values were compared
with the manual procedure and square mean deviation shows that the average value is the closest.
Combining the binary mask with the intensity image allows determining the pressure distribution on
foot plant. Some descriptors like maximum pressure and mean pressure were calculated for each
sample. The ratio between them measures the uniformity of the distribution and is independent of foot
size and person’s weight. Finally, pressure distribution and morphological classification of feet were
compared to show the existence of high correlation between them.
The approach was methodologically appropriate for a preliminary work. As a future work we plan to
build a protocol to evaluate it with a larger number of patients in order to validate clinically this
technique.

6. References
[1] Revenga C and Bulo M 2005 El pie plano valgo: evolución de la huella plantar y factores
relacionados Revista Española de Cirugía Ortopédica y Traumatología 49 271-280
[2] Buchelly F, Mayorca D, Ramírez A, Mayorca H and Ferrín D 2009 Elaboración de un
podobarógrafo para análisis de la huella plantar Fundación Universitaria María Cano
(Popayán, Colombia: Proyecto de la asignatura de biomecánica para el programa de
fisioterapia) p 1-32
[3] Oller A 2006 La fórmula metatarsial y su valor predictivo en los trastornos de la marcha
Universitat de Barcelona (Barcelona: Tesis doctoral del programa de Doctorado Investigació
en Ciències de la Salut) p 1-130
[4] Coll M, Viladot A and Suso A 1999 Estudio evolutivo del pie plano infantil Revista española de
cirugía ortopédica y traumatología 43 129-136
[5] Mayorca D 2011 Diseño y construcción de un sistema óptico de análisis de distribución de
presiones plantares para el diagnóstico de pie diabético mediante técnicas de visión por
computador Universidad del Cauca (Popayán, Colombia: Trabajo de grado de la carrera de
Ingeniería Física) p 1-74
[6] Vélez M 2009 “Posturología como análisis preventivo de lesiones músculo-esqueléticas.”
Empresa Preven-Ergo (Quito-Ecuador) 1-5
[7] Aguilar I, Sánchez I, Pedraza G and Guadarrama L 2012 Correlación plantar y maloclusión.
Caso clínico. Revista ADM 49 91-94
[8] Pinzón M Podología Unos pies sanos son la base de un cuerpo sano BENAmedic clínica web
page: http://www.clinicabenamedic.com/podologia/ Access date: 05-06-2015
[9] Lara S, Lara A, Zagalaz M and Martínez E 2011 Análisis de los diferentes métodos de
evaluación de la huella plantar Retos. Nuevas tendencias en Educación Física, Deporte y
Recreación 19 49-53
[10] Berdejo D, Lara A, Martínez E, Chacón J and Lara D 2013 Footprint modifications according to
the physical activity practised Rev.int.med.cienc.act.fìs.deporte vol. 13 num. 49 19-38
[11] Ferrín D, Magdalena X and Loaiza H 2013 Determinación semiautomática de parámetros
morfológicos de la huella plantar mediante el procesamiento digital de imágenes Revista
S&T 11(27) 9-26
[12] Arcan M and Brull M 1976 A fundamental characteristic of the human body and foot. The foot-
ground pressure pattern J. Biomechanics 9 453-457

9
SABI 2015 IOP Publishing
Journal of Physics: Conference Series 705 (2016) 012020 doi:10.1088/1742-6596/705/1/012020

[13] Lord M 1981 Foot pressure measurement: A review of methodology Journal of Biomedical
Engineering 3 91-99
[14] Bates K, Savage R, Pataky T, Morse S, Webster E, Falkingham P, Ren L, Qian Z, Collins D,
Bennett M, McClymont J and Crompton R 2013 Does footprint correlate with foot motion
and pressure? Interface. Journal of the royal society 10(83)
[15] Arce C 2015 Arces Software Dr. Carlos Alberto Arce Gonzales, Medio especialista en medicina
de rehabilitación http://www.arcesw.com/principal.htm Acces date: 05-07-2015
[16] Salazar C 2007 Pie plano, como origen de alteraciones biomédicas en cadena ascendente
Fisioterapia 29(2) 80-89
[17] Curvale G and Rochwerger A 2002 Encyclopédie Medico-Chirurgicale E-14-429 Subject title:
Pie cavo. Editions Scientifiques et Médicales Elsevier SAS (París)
[18] Zurita F, Martínez A and Zurita A 2006 Influencia de la tipología del pie en la actividad físico
deportiva Fisioterapia 29(2) 74-79
[19] López G and Coloma S 2012 Hallux valgius y presión plantar. Revisión de la literatura I
simposio sobre biomecánica y pie durante la actividad física – Universitat de Valencia (3-4
de Mayo de 2012)
[20] Tizón E, Dovale M, Fernández C, López M, Mouteira M, Penabad S, Rodríguez O and Vásquez
R 2004 Atención de enfermería en la prevención y cuidados del pie diabético Aten Primaria
34(5) 263-271
[21] Tazi O and Debure C 2011 Encyclopédie Medico-Chirurgicale E – 98-866-B-10 Subject title: Pie
diabético. Editions Scientifiques et Médicales Elsevier SAS (Paris)
[22] Gómez E, Levy A, Díaz A, Cuesta M, Montañez C and Calle A 2012 Pie diabético Seminarios
de la fundación española de Reumatología 13(4) 119-129
[23] Díaz C, Torres A, Ramírez J, García L and Álvarez N 2006 Descripción de un sistema para la
medición de las presiones plantares por medio del procesamiento de imágenes. Fase I
Revista EIA 6 43-55
[24] Shah S and Patil K 2005 Processing of foot pressure images and display of an advanced clinical
parameter PR in Diabetic Neuropathy Proceedings of the 2nd international IEEE EMBS
conference on Neural Engineering Arlington, Virginia March 16-19, 2005 p v-viii
[25] Periyasamy R, Mishra A, Anand S and Ammini A 2010 Static foot pressure image analysis for
variation in men and women while standing. Foot Pressure image analysis Proceedings of
2010 international conference on systems in medicine and biology 16-18 December 2010,
IIT Kharagpur, India p 386-391
[26] Szeliski R 2010 Computer Vision: Algorithms and Applications Springer p 327-342
[27] Ferrín C and Buchelly F 2010 Diseño y construcción de plataformas multitoque de tipo óptico,
utilizando técnicas de visión artificial en tiempo real, para facilitar la interacción de los
usuarios con el computador Universidad del Cauca (Popayán, Colombia: Trabajo de grado
de la carrera de Ingeniería Física) p 28-31
[28] 2009 Opencv 2.0 Documentation p 313
[29] Gonzalez R and Woods R Digital image processing (Second Edition) Prentice Hall (Upper
Saddle River, New Jersey) p 528-531

Acknowledgments
Authors acknowledge the assistance from the GOL members (Grupo de Optica y Láser) of
Universidad Del Cauca, especially to engineers Leonairo Pencue Fierro and Carlos Diego Ferrín.

10