You are on page 1of 7

LWT - Food Science and Technology 59 (2014) 1220e1226

Contents lists available at ScienceDirect

LWT - Food Science and Technology


journal homepage: www.elsevier.com/locate/lwt

Measurement parameter of color on yacon (Smallanthus sonchifolius)


slices using a computer vision system
~ a a, Raúl Siche b, *, Wilson Castro c, Rosmer Huama
Erick Saldan n d, Roberto Quevedo e
a ~o Paulo, Piracicaba City, SP 13418-900, Brazil
Department of Agro-industry, Food and Nutrition, “Luiz de Queiroz” Agricultural College, University of Sa
b
Facultad de Ciencias Agropecuarias, Universidad Nacional de Trujillo, Av. Juan Pablo II s/n. Ciudad Universitaria, Trujillo, Peru
c
Facultad de Ingeniería y Ciencias Agrarias, Universidad Nacional Toribio Rodríguez de Mendoza de Amazonas, Pampas de Higos Urco S/N, Ciudad
Universitaria, Amazonas, Peru
d
Escola de Química e Alimentos, Universidade Federal de Rio Grande, Av. Italia km 8, Campos Carreiros s/n. Rio Grande, Brazil
e
Food Science and Technology Department, Universidad de Los Lagos, Osorno, Chile

a r t i c l e i n f o a b s t r a c t

Article history: The aim of this research was to design, implement and calibrate a Computer Vision System (CVS), for use
Received 28 January 2014 in real-time, in order to measure the color on minimally processed yacon slices. For this purpose, a device
Received in revised form (software and hardware) was designed and implemented which consisted of two steps: a) image
4 June 2014
acquisition and b) image processing and analysis. For both, an algorithm and a graphical user interface
Accepted 16 June 2014
Available online 24 June 2014
(GUI) were developed in MatLab. CVS calibration was performed with a conventional colorimeter (Model
CIE L*a*b*). Minimally processed yacon slices were obtained and stored at 5  C. Color changes were
estimated every 2 h, for 26 h, obtaining its color parameters. L* decreased from 65.9 to 60.8, with a
Keywords:
CIE L*a*b* color space
tendency to a black color; a* increased from 7.3 to 17.7, approaching a red color; b* increased from 35.1 to
Minimally processed food 41.5, presenting a tendency to a yellow color. Moreover, C* increased from 35.9 to 40.1, H* decreased from
Color system 78.2 to 66.9 and DE* increased from 2.6 to 13.2. Low errors calculated (eL* ¼ 5.001%, and ea* ¼ 2.287%, and
Yacon eb* ¼ 4.314%) ensure suitable and efficient application in industrial process automation and quality
control in the food industry.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction Color is one of the fundamental parameters defining the quality


of a food. The assessment of this sensory parameter is crucial and is
The focus on nutrition and health of the populations in Western a concern in the food industry (Du & Sun, 2004). In this context,
countries caused a great interest in the identification of new sensory properties of food can be estimated to the appearance and
functional foods and ingredients to prevent diseases such as dia- color of the surface, which are the first visually evaluated param-
betes, cancer, etc. (Yan et al., 1999). South America is a true treasure; eters and consequently consumer relationship in the acceptance or
a source of a wide range of fruits and vegetables that are clearly rejection of the product even before it enters the mouth (Pedreschi,
underutilized. One of these rediscovered treasures is the yacon, an Leon, Mery, & Moyano, 2006).
Andean crop that grows at altitudes of 1000e3200 m.a.s.l., and was Currently, new tools are being used to measure changes in the
domesticated in the Andes by pre-Incacivilizations. The centers of color characteristics of food. One of these being the computer vision
greater diversity are located in the borderlands of Peru with Bolivia technique projected as an alternative to the sensory evaluation. The
and Ecuador (Delgado, Tamashiro, Marostica, & Pastore, 2013). The efficient use of computer vision techniques for food color and
interest by said root has focused in its anti-diabetic effect because it assessment requires a calibration process based on absolute com-
modulates insulin concentration in blood plasma and inhibits the mon colors for color data as knowledge of the image characteristics
hepatic gluconeogenesis (Genta et al., 2009). In addition to these that could correlate better with the quality product (Mendoza,
health benefits, it has been reported to have antioxidant effects Dejmek, & Aguilera, 2006).
(Takenaka et al., 2003). The computer vision is in relationship with the image process-
ing and analysis, allowing the color segmentation, quantification
and classification of elements of interest. The utilization of com-
* Corresponding author. Tel.: þ51 44294778.
puter vision as an instrument for estimating the changes of the
E-mail addresses: rsiche@unitru.edu.pe, raul_siche@hotmail.com (R. Siche). color characteristics on food is extremely advantageous when it is

http://dx.doi.org/10.1016/j.lwt.2014.06.037
0023-6438/© 2014 Elsevier Ltd. All rights reserved.
~ a et al. / LWT - Food Science and Technology 59 (2014) 1220e1226
E. Saldan 1221

used for replace the panellists responses for images taken with a connected to a computer Core i3, vertical position at 22.5 cm from
video camera and after processed and analyzed by computer al- the sample and an angle of 45 with the light source) (Fig. 1). The
gorithms (Saldan ~ a, Siche, Lujan, & Quevedo, 2013). Doing this re- lamps were arranged 35 cm squared above the sample and at an
duces the cost, time and subjectivity of color measurements, angle of 45 with the sample plane to give a uniform light intensity
making the estimation more reliable and scientific (Brosnan & Sun, over the food sample. In addition, the surface where the sample is
2004). located is natural rubber (black color).
The combining of a digital camera and its image processing
software that replaces the traditional measuremental instruments 2.2. CVS implementation
have been widely used to provide a cheaper and versatile form to
measure the color of many foods (Blasco, Aleixos, & Molto, 2007; The extraction of the color characteristics of the imaging was
Mendoza et al., 2006; Pedreschi et al., 2006). performed on a graphical user interface (GUI) created in MatLab
Few studies on inspection of processed fruit apply computer described below.
visuals, likely due to its lower economic interest in respect to fresh Yacon samples were conditioned in the CVS, the lights were
fruit and the greater complexity of the processing of these. How- turned on, waiting 10 min so the light is uniformly distributed
ever, the status of current inspection systems opens new possibil- within the system, and then the GUI was started. Followed by im-
ities for the creation of complex and robust algorithms capable of age acquisition and segmentation by thresholding to finally get the
running in real-time inspection lines. Therefore, it is important to color parameters L*a*b* from the RGB parameters using equations.
start to develop appropriate decision algorithms and methods that The first phase involves the conversion from RGB to XYZ (Poynton,
allow acceleration and optimize industrial processes, particularly in 1996) (Eqs. (1)e(7)) and the second of XYZ to CIE Lab color space
the case of fruit. (Eqs. (8)e(13)).
In this regard, it is of interest to use digital video cameras with
communication to computers for the color valuation in real-time If R=255 > 0:04045/r ¼ ððR þ 0:055Þ=1:055ÞÞ2:4 ;
allowing the processing and analysis of their images with the use (1)
if R=255  0:04045/r ¼ R=12:92
of special programs, so that the integrated system interprets in-
formation of manner equivalent to the human eye. Measurements
of the number, length, width, area, color, etc. of a particular food If G=255 > 0:04045/g ¼ ððG þ 0:055Þ=1:055ÞÞ2:4 ;
(2)
would be trivial. Thus, the aim of this research was to design, if G=255  0:04045/g ¼ G=12:92
implement and calibrate a computer vision system (CVS), for use in
real-time, in order to measure color on minimally processed yacon
If B=255 > 0:04045/b ¼ ððB þ 0:055Þ=1:055ÞÞ2:4 ;
slices. (3)
if B=255  0:04045/b ¼ B=12:92
2. Materials and methods Subsequently, these values were converted of rgb values to XYZ
using the matrix M for a D65-2 illuminant-observer (Eq. (4)). The
2.1. CVS design D65-2 illuminant-observer is a standard recommended by the CIE
and best suited to the lighting that was used (Blasco et al., 2007).
The CVS consisted of a wood box whose internal walls were
2 3 2 3
painted black to avoid the light and reflection from the room X r
(Pedreschi et al., 2006), illuminated by four fluorescent lights 4 Y 5 ¼ 100*½M 4 g 5 (4)
(Philips Master TL-D Super 80, natural daylight, 18 W, length 60 cm, Z b
2 3
color temperature of 6500 K, color rendering index close to 95%) 0:4124 0:3576 0:1805
and the capture system (Sony Handycam HDR-CX11E video camera, where ½M ¼ 4 0:2126 0:7152 0:0722 5
0:0193 0:1192 0:9505

* *
a and b
*
Fig. 1. Scheme for converting RGB images to units L* a* b* using CVS (b
L ,b b are estimated values).
1222 ~ a et al. / LWT - Food Science and Technology 59 (2014) 1220e1226
E. Saldan

Using the M matrix, are obtained X, Y and Z (Eqs. (5)e(7)). where P: parameters were obtained with colorimeter; parameters
were obtained with CVS; DP: DL ¼ 100, Da ¼ Db ¼ 240 (consid-
X ¼ 100*r*0:4124 þ 100*g*0:3572 þ 100*b*0:18050 (5) ering that 0  L*  100, 120  a*  þ120
and 120  b*  þ120); and N: 90 (number of measurements). To
Y ¼ 100*r*0:2126 þ 100*g*0:7152 þ 100*b*0:0722 (6) evaluate the performance of the model used, the overall error (e)
was calculated (Eq. (18)).
Z ¼ 100*r*0:0193 þ 100*g*0:1192 þ 100*b*0:9505 (7)
e ¼ ðeL* þ ea* þ eb* Þ=3 (18)
Subsequently, x, y, z was obtained (x ¼ X/Xn; y ¼ Y/Yn and z ¼ Z/
Zn), where Xn ¼ 95.047, Yn ¼ 100 and Zn ¼ 108.883, are tri-stimulus 2.4. Raw material and minimally processed yacon
values obtained by the weighted-ordinate method (Dl ¼ 1 nm),
with the CIE 1964 standard observer (10 visual field) and the CIE Yacon roots, from Cajamarca (Peru), were selected and classified
standard illuminant D65, as references (CIE, 1995). by size and quality (approximately 200 g of cylindrical shape),
After, varX, varY and varZ were calculated with Eqs. (8)e(10). eliminating defective roots. The roots were washed with tap water
and neutral detergent in order to remove traces of dust and dirt,
If x > 0:008856/varX ¼ ðxÞ1=3 ; if x  0:008856/varX and these were immersed in a sodium hypochlorite solution
¼ ð7:787*xÞ þ ð16=116Þ (100 ppm) for three minutes in order to reduce its microbial load.
To reduce metabolic activity, the roots were stored at 12  C for 18 h.
(8)
The roots were peeled and sliced (thickness: 20 mm), then the
If y > 0:008856/varY ¼ ðyÞ1=3 ; if y  0:008856/varY slices were immersed in a sodium hypochlorite solution (100 ppm)
¼ ð7:787*yÞ þ ð16=116Þ for three minutes and immediately drained for two minutes. After
the slices were immersed in an acid solution (1% citric acid) for
(9)
2 min and drained for 2 min. Finally the samples were packed in
If z > 0:008856/varZ ¼ ðzÞ1=3 ; if z  0:008856/varZ bags of polyethylene terephthalate (PET) with polyvinyl chloride
(PVC) film and stored at 5  C.
¼ ð7:787*zÞ þ ð16=116Þ
(10) 2.5. Measurement parameter of color on yacon slices
Subsequently, L*, a* and b* were calculated (Eqs. (11)e(13)).
The color parameters (L*, a*, b*, C*, H* and DE*) were measured
*
L ¼ ð116*varYÞ  16 (11) in three samples of yacon slices (stored at 5  C), every 2 h for a
period of 26 h. Once these samples were acquired and processed,
a* ¼ 500*ðvarX  varYÞ (12)
the images were evaluated (720  480 pixels) in rectangular co-
ordinates (L*, a*, b*). The L* parameter (luminosity) is an attribute
b* ¼ 200*ðvarY  varZÞ (13) by which a surface emits more or less light and can take values
Color measurements were expressed in cylindrical coordinates between 0 (absolute black) to 100 (absolute white). The parame-
C* and H* (C* defines the chromaticity and H* indicates the hue) ters a* and b* represent the chromaticity, where a* defines the
(Eqs. (14) and (15)) (McLaren, 1976). red-green component (red for positive values and green for
negative values) and the b* parameter defines the yellow-blue
 2  2 1=2 component (yellow for positive values and blue for negative
C* ¼ a* þ b* (14)  n et al., 2006). Finally, using SPSS 15.0 for Windows
values) (Leo
(SPSS Inc. Chicago, USA), these parameters were plotted against
  
H* ¼ arctan b* a* (15) time.

To quantify the total color changes, the total color difference was 3. Results and discussion
calculated (DE*) between minimally processed yacon slices (stored
at 5  C) for a time of 2e26 h (every 2 h) and minimally processed 3.1. CVS design
yacon slices to initial condition (time: 0) (Eq. (16)).
 The lighting condition is very important for image acquisition.
2  2  2 1=2
DE* ¼ Da* þ Db* þ DL* (16) Lighting was positioned vertically with an angle of 45 , thereby
preventing direct reflections to the camera when the sample
2.3. CVS calibration plane is illuminated. Fernandez, Castillero, and Aguilera (2005)
used this position to illuminate apple slices in where they were
In order to calibrate, the CVS were measured by 30 color charts controlled and its dehydration evaluated by the tracking of its
(with three replicates) which obtained the values of L*a*b* with color and shape. A similar system later was used by Pedreschi
Konika Minolta cr400 Colorimeter in parallel with the same CVS; et al. (2006) to illuminate potato chips in a system designed to
this subsequently measured the error of CVS in respect to the measure the kinetics of color changes under different frying
colorimeter. The mean normalized error for each parameter (P: L*, temperatures.
a* and b*) by comparing with model estimates was obtained with In the CIE L*a*b* color system used in this experiment, X, Y and Z
 n, Mery, Pedreschi, & Leo
the Eq. (17) (Leo n, 2006). depend on the geometric measurement, illumination and cameras
  based in charge-coupled devices (CCD); therefore, if any of these
 ∧  parameters changed, L*, a* and b* also changed. On the other hand,
N Pi
  Pi 
1 X each device produces its own RGB values for the same pixel in a
eP ¼ (17)
N DP scene. For this reason, attempts they have been developed for
i¼1
standardization of values known as the called RGB color space. A
color space used in food inspection, nearest to the human color
~ a et al. / LWT - Food Science and Technology 59 (2014) 1220e1226
E. Saldan 1223

perception is the Hue, Saturation and Intensity (HSI). Zou, Zhao, and 3.2. CVS implementation
Li (2007) used both RGB and HSI for Fuji apples, and classified their
findings into four color categories. Abdullah, Mohamad-Saleh, The implementation was done using a digital video camera
Fathinul-Syahir, and Mohd-Azemi (2006) converted RGV co- connected to a computer that governs the entire CVS. Typically
ordinates into HSI coordinates, where the H* component was used video cameras operate in the RGB space. The processing of each
to classify starfruit (Averrhoa carambola) in four maturity cate- picture begins with the segmentation process. Image segmentation
gories. However, both RGB and HSI color spaces are not uniform. ratio of green and blue (G/B) was applied to each pixel, where a
This means that the Euclidean distance between two points located pixel with a value above of the threshold value was considered the
in different color regions in these areas do not produce the same sample and segmented from the background. This relationship was
difference of perception as the standard observer. To try to solve established because the analyzed samples (corresponding to
this problem, we defined uniform spaces as the CIE Lab and Hunter minimally processed yacon slices) have very low blue values but
Lab, which are usually used for the color comparison (Leo n et al., high green values, while the background always had lower values
2006). Mendoza et al. (2006) compared different color spaces facilitating discrimination. The pixels were converted to L*a*b*
such as sRGB, HSV and L* a* b* in terms of their suitability for color space according to equations (11)e(13). This step represented a loss
quantization on curved surfaces, proving to be more appropriate L* of precision because the RGB space is limited in the number of
a* b*. colors (limited spectrum). Moreover, there are many arrays for

Table 1
CVS Programming in MatLab.

Step MatLab command Description

1 Im ¼ imread ('name.format'); Read the RGB image and stores it in the array im.
2 img ¼ rgb2gray (im); Change the image to grayscale and stores it in img.
3 [NFIL ncol] ¼ size (img); Determines the size of the image.
4 for i ¼ 1: NFIL, for j ¼ 1: ncol, if img (i, j) < 85, img (i, j) ¼ 250, end, end, end Change the background color.
5 Th ¼ 0.90; Determine the threshold (Th), for this is used imtool.
6 imbw ¼ im2bw (img, Th); Binarize the image using Th and is stored into imbw matrix.
7 IMGC ¼ imcomplement (imbw); Generate the complement of the binarized image.
8 IMGC ¼ imfill (IMGC, 'holes'); Fill existing holes.
9 [Imge n] ¼ bwlabel (IMGC); Tag and stored the image in the image matrix.
10 stats ¼ regionprops (imge, 'all'); Gets statistical information and stores it in the stats array.
11 Area ¼ zeros (1, n); Determine the area.
12 for i ¼ 1: n; Area (i) ¼ stats (i).Area; end Create zeros matrix.
k ¼ 0, for i ¼ 1: n, if max (Area) ¼¼ stats (i).Area k ¼ i;
13 Cn ¼ stats (k).Centroid; Get the centroid.
14 Deq ¼ stats (k).EquivDiameter; Gets the equivalent diameter.
hold on, plot (Cn (1), Cn (2), '* r'), hold off; break% stop process, end, end
15 POSP ¼ stats (k).PixelList; % save POSP POSP; [NFIL ncol] ¼ size (PPD), for i ¼ 1: Obtains information about the image position.
NFIL, x (i) ¼ POSP (i, 1), and (i) ¼ POSP (i, 2);
16 R (i) ¼ Im (y (i), x (i), 1), G (i) ¼ Im (y (i), x (i), 2), B (i) ¼ Im (y (i), x (i), 3); end Obtains values of R, G, B.
17 image (label2rgb (imge)); Displays the new image labeled.
18 IMAR ¼ double (R)/255; Save the image in the matrices labeled: RR, GG, B B.
imag ¼ double (G)/255;
IMAB ¼ double (B)/255;
19 [NFIL ncol] ¼ size (IMAR); Determines the matrix size and conditions the first part
for i ¼ 1: NFIL, for j ¼ 1: ncol if (IMAR (i, j) > 0.04045); IMAR (i, j) ¼ ; of the conversion from RGB to XYZ.
((IMAR (i, j) þ 0.055)/1055)^2.4; else
; IMAR (i, j) ¼ IMAR (i, j)/12.92;
end, end, end, for i ¼ 1: NFIL, for j ¼ 1: ncol if (imag (i, j) > 0.04045),
imag (i, j) ¼ ((imag (i, j) þ 0055)/1.055)^2.4; else
Imag (i, j) ¼ imag (i, j)/12.92; end; end end; for i ¼ 1: NFIL, for j ¼ 1:
ncol if (IMAB (i, j) > 0.04045); IMAB (i, j) ¼ ; ((IMAB (i, j) þ 0055)/1055)^2.4; else;
IMAB (i, j) ¼ IMAB (i, j)/12.92; end; end; end; IMAR ¼ Imar* 100; imag ¼ imag* 100;
IMAB ¼ IMAB* 100;
20 X ¼ IMAR.* 0.4124 þ Imag.* 0.3576 þ IMAB.* 0.1805; Performs the second part of the transformation from
Y ¼ IMAR.* 0.2126 þ Imag.* 0.7152 þ IMAB.* 0.0722; RGB to XYZ, constants for the Observer ¼ 2 , Illuminant ¼ D65
Z ¼ IMAR.* 0.0193 þ Imag.* 0.1192 þ IMAB.* 0.9505;
21 var_X ¼ X./95.047; Transforms from XYZ to Var_X, var_Y, var_Z, using
var_Y ¼ Y./100.000; Observer ¼ 2 , Illuminant ¼ D65
var_Z ¼ Z./108.883;
22 [NFIL ncol] ¼ size (var_X); Determine the matrix dimension
23 for i ¼ 1: NFIL, for j ¼ 1: ncol; if; (var_X (i, j) > 0.008856); var_X Determine the values: var_X (i, j), var_Y (i, j), var_Z (i, j)
(i, j) ¼ var_X (i, j).^(1/3) else
var_X (i, j) ¼ (7787* var_X (i, j)) þ (16/116);
end; endn end; for i ¼ 1: NFIL, for j ¼ 1: ncol if (var_Y (i, j) > 0.008856);
var_Y (i, j) ¼ var_Y (i, j).^(1/3); else
var_Y (i, j) ¼ (7787* var_Y (i, j)) þ (16/116);
end, end, end, for i ¼ 1: NFIL, for j ¼ 1: ncol
if (var_Z (i, j) > 0.008856); var_Z (i, j) ¼ var_Z (i, j).^(1/3)
else var_Z (i, j) ¼ (7787* var_Z (i, j)) þ (16/116);
end, end, end
24 L ¼ (116* var_Y)16; Obtain L*, a*, b* values
a ¼ 500* (var_Xvar_Y);
b ¼ 200* (var_Yvar_Z);
1224 ~ a et al. / LWT - Food Science and Technology 59 (2014) 1220e1226
E. Saldan

conversion from RGB to XYZ. This is because there is a dependency


of the RGB working space in many algorithmic software (Adobe
RGB, Apple RGB, CIE RGB, RGB NTSC, PAL/SECAM RGB, sRGB, etc.).

3.3. Developed algorithm for image processing

An image analysis algorithm made in MatLab (Table 1) was


implemented in order to estimate the color and a GUI simple to
characterize the color on the flat surfaces of food.

3.4. CVS calibration

The CVS is designed to display the color as a result and this is


reproducible at any time. Small differences were observed between
the values of the color parameters (L*, a* and b*) obtained by the
colorimeter and the CVS (Fig. 2). It was also observed that the
model conversion from RGB to L*a*b*, obtained from images
generated CVS is reliable. Fig. 2 shows the relationship between the Fig. 3. Normalized frequency distribution of the colorimetric parameters estimated by
two devices, with relatively low error (eL* ¼ 5.001%, ea* ¼ 2.287% CVS [L* (A), a* (:), b* (C)]. The intervals show the mean ± standard errors.
and eb* ¼ 4.314%). An error less than 10% indicates that the model
explains the experimental data acceptably (Gutie rrez-Pulido & de
la Vara Salazar, 2004). Also observed was a correlation coefficient giving an idea of the colorimetric evolution experienced by the
near one (r ¼ 0.9982), indicating a high degree of correlation. sample. The result of enzymatic pigmentation causes to a wide
range of L*, a* and b* values (Nicolas, Richard-Forget, Goupy, Amiot,
3.5. Estimation and evaluation of the colorimetric parameters & Aubert, 1994; Robards, Prenzler, Tucker, Swatsitang, & Glover,
1999), being a* suitable for assessing change of color from green
Color is an important property related to the chemical compo- to red in fruits (Castaner, Gil, Artes, & TomasBarberan, 1996). Fig. 4
sition and the degree of alteration of food over time. The different shows that most of the a* values are positive, indicating a tendency
values of Luminosity (L*) indicates that the reflection of the light in toward red. In addition, a* change from 7.33 (0 h) to 17.70 (26 h),
a yacon slice was heterogeneous (Fig. 3). This may have occurred as also interpreted as an approximation to red. The variation of the b*
a result of enzymatic browning, whereby polyphenoloxidase (PPO) was not as significant as for the a*, registering an insignificant in-
catalyzes the conversion of phenolic compounds to quinones, crease with exposure time (from 35.11 to 41.49, for 0 and 26 h
which are polymerized to form dark colored pigments (Sapers & respectively) and a trend toward yellow. These results are consis-
Douglas, 1987). When some fruit is cut, such as yacon, the cut tent with data obtained by Krokida, Oreopoulou, Maroulis, and
surface turns dark, reducing not only the visual quality, but also Marinos-Kouris (2001) and Segnini, Dejmek, and Oste (1999).
results in undesirable changes in the flavor and nutrient loss (Luo & From L*, a* and b* psychometric coordinates H* and C* were
BarbosaCanovas, 1997). obtained. H* is the hue angle measurement and indicates the
L* decreases as the samples are darker. Fig. 4 shows that the relative orientation of the color with respect to the origin (0 )
samples are decreasing in value L* versus time. For example, (Chiralt, Martínez-Navarrete, Gonz alez, Talens, & Moraga, 2007).
average L* went from 65.99 (0 h) to 60.84 (26 h), meaning that the H* decreases from 78.22 to 66.90 in the 26 h of treatment
samples darkened. (Fig. 5). These results indicate that the variation in the yellow-green
The parameters a* and b* determine a plane which can repre- quadrant. C*, or chroma, is the measure of the saturation of the
sent the individual and combined changes in both parameters, color being measured from the center point of the color space, i.e.,

Fig. 2. Comparison among colorimetric parameters obtained by CVS and colorimetric


parameters obtained by colorimeter [L* (A), a* (:), b* (C)]. The intervals show the Fig. 4. Variation of the colorimetric parameters estimated by CVS [L* (A), a* (:), b*
mean ± standard errors. (C)] in function of time. The intervals show the mean ± standard errors.
~ a et al. / LWT - Food Science and Technology 59 (2014) 1220e1226
E. Saldan 1225

between those that have and do not have the required minimum
quality.
In future work, the computer vision system should be integrated
with an artificial neural network using colorimetric and morpho-
metric parameters of food. This integrated system would classify
processed or fresh foods according to their sensory quality, and
could even be used to estimate sensory shelf life of minimally
processed foods.

Acknowledgment

Our acknowledgment to Mrs. Annie Escobedo for his valuable


review of the English language.

References
Fig. 5. Variation of H* (-), C* ( ) and DE* (▫) with time. The intervals show the
mean ± standard errors.
Abdullah, M. Z., Mohamad-Saleh, J., Fathinul-Syahir, A. S., & Mohd-Azemi, B. M. N.
(2006). Discrimination and classification of fresh-cut starfruits (Averrhoa
carambola L.) using automated machine vision system. Journal of Food Engi-
the achromatic axis where a* ¼ b* ¼ 0, and along a straight line neering, 76(4), 506e523. http://dx.doi.org/10.1016/j.jfoodeng.2005.05.053.
extending outwardly from that point (Chiralt et al., 2007). The Baysal, T., Icier, F., Ersus, S., & Yıldız, H. (2003). Effects of microwave and infrared
drying on the quality of carrot and garlic. European Food Research and Tech-
change in color parameters a* and b* lead to a variation of the nology, 218(1), 68e73.
parameter C* of the samples analyzed (Nicolas et al., 1994; Sapers & Blasco, J., Aleixos, N., & Molto, E. (2007). Computer vision detection of peel defects
Douglas, 1987), this being one of the major factors limiting the in citrus by means of a region oriented segmentation algorithm. Journal of Food
Engineering, 81(3), 535e543. http://dx.doi.org/10.1016/j.jfoodeng.2006.12.007.
useful life of minimally processed products. C* value increases from Brosnan, T., & Sun, D. W. (2004). Improving quality inspection of food products by
35.87 to 45.11 in the 26 h of treatment, i.e. yacon slices become computer vision - a review. Journal of Food Engineering, 61(1), 3e16. http://
more chromatic, with a tendency towards a combination of red- dx.doi.org/10.1016/s0260-8774(03)00183-3.
Castaner, M., Gil, M. I., Artes, F., & TomasBarberan, F. A. (1996). Inhibition of
yellow. browning of harvested head lettuce. Journal of Food Science, 61(2), 314e316.
The color total difference (DE*) in CIE L*a*b* is the distance http://dx.doi.org/10.1111/j.1365-2621.1996.tb14184.x.
between the locations of color in the CIE. In recent studies, the total Chiralt, A., Martínez-Navarrete, N., Gonz alez, C., Talens, P., & Moraga, G. (2007).
Propiedades físicas de los alimentos. Espan ~ a: Editorial Universidad Politecnica de
color difference (DE*) has been used in the determination of the Valencia.
color change in various foods during processing and storage: Choi, L. H., & Nielsen, S. S. (2005). The effects of thermal and nonthermal processing
thermal and non-thermal processing of apple cider (Choi & Nielsen, methods on apple cider quality and consumer acceptability. Journal of Food
Quality, 28(1), 13e29.
2005), dried carrot and garlic using microwave and infrared (Baysal,
CIE. (1995). Industrial colour-difference evaluation (Technical Report). Publication CIE
Icier, Ersus, & Yıldız, 2003), dried kiwi with hot air and microwave No. 116. Vienna: Central Bureau of the Commission Internationale de
(Maskan, 2001), among others. In this research, the overall color L'Eclairage.
changes (DE*) of minimally processed yacon slices increased from Delgado, G. T. C., Tamashiro, W., Marostica, M. R., & Pastore, G. M. (2013). Yacon
(Smallanthus sonchifolius): a functional food. Plant Foods for Human Nutrition,
2.57 to 13.23, showing a clear color value trend of: red, yellow and 68(3), 222e228. http://dx.doi.org/10.1007/s11130-013-0362-0.
dark. Du, C. J., & Sun, D. W. (2004). Recent developments in the applications of image
processing techniques for food quality evaluation. Trends in Food Science &
Technology, 15(5), 230e249. http://dx.doi.org/10.1016/j.tifs.2003.10.006.
Fernandez, L., Castillero, C., & Aguilera, J. M. (2005). An application of image analysis
to dehydration of apple discs. Journal of Food Engineering, 67(1e2), 185e193.
4. Conclusions http://dx.doi.org/10.1016/j.jfoodeng.2004.05.070.
Genta, S., Cabrera, W., Habib, N., Pons, J., Carillo, I. M., Grau, A., et al. (2009). Yacon
syrup: Beneficial effects on obesity and insulin resistance in humans. Clinical
The contribution of this study was to develop a new algorithm Nutrition, 28(2), 182e187. http://dx.doi.org/10.1016/j.clnu.2009.01.013.
using computer vision and apply this methodology to yacon slices, Gutierrez-Pulido, H., & de la Vara Salazar, R. (2004). Ana lisis y disen
~ o de exper-
imentos. Me xico DF: McGraw-Hill.
which is an unexploited functional fruit. This algorithm provides
Krokida, M., Oreopoulou, V., Maroulis, Z., & Marinos-Kouris, D. (2001). Colour
objective sensory information about color. Therefore, this proposal changes during deep fat frying. Journal of Food Engineering, 48(3), 219e225.
could take charge over traditional systems based on subjective  n, K., Mery, D., Pedreschi, F., & Leo
Leo n, J. (2006). Color measurement in L*a*b*
color measurements like sensory panel evaluation, where influence units from RGB digital images. Food Research International, 39(10), 1084e1091.
http://dx.doi.org/10.1016/j.foodres.2006.03.006.
aspects such as fatigue, stress, level of vision, among others are Luo, Y., & BarbosaCanovas, G. V. (1997). Enzymatic browning and its inhibition in
evident. new apple cultivars slices using 4-hexylresorcinol in combination with ascorbic
The algorithm implemented in the CVS enabled an estimate of acid. Food Science and Technology International, 3(3), 195e201. http://dx.doi.org/
10.1177/108201329700300307.
the parameters of CIE Lab color with low errors (eL* ¼ 5.001%, Maskan, M. (2001). Kinetics of colour change of kiwifruits during hot air and mi-
ea* ¼ 2.287% and eb* ¼ 4.314%). L* decreased from 65.99 to 60.84 crowave drying. Journal of Food Engineering, 48(2), 169e175.
due to obscuration of the samples, a* increased from 7.33 to 17.70 McLaren, K. (1976). The development of CIE (L, a, b) uniform color space and color-
difference formula. Journal of the Society of Dryers and Colourist, 92, 339e341.
indicating a red color trend, and b* increased from 35.11 to 41.49 Mendoza, F., Dejmek, P., & Aguilera, J. M. (2006). Calibrated color measurements of
suggesting a yellow color trend with time. C* increased from 35.87 agricultural foods using image analysis. Postharvest Biology and Technology,
to 45.11, showing a tendency towards a combination of red-yellow. 41(3), 285e295. http://dx.doi.org/10.1016/j.postharvbio.2006.04.004.
Nicolas, J. J., Richard-Forget, F. C., Goupy, P. M., Amiot, M. J., & Aubert, S. Y. (1994).
H* declined from 78.22 to 66.90, indicating a variation in the
Enzymatic browning reactions in apple and apple products. Critical Reviews in
yellow-green quadrant. DE* increased from 2.57 to 13.23, showing a Food Science & Nutrition, 34(2), 109e157.
tendency to colors: red, yellow and dark. These results demonstrate Pedreschi, F., Leon, J., Mery, D., & Moyano, P. (2006). Development of a computer
that the designed system based on image helps explain the color vision system to measure the color of potato chips. Food Research International,
39(10), 1092e1098. http://dx.doi.org/10.1016/j.foodres.2006.03.009.
changes of minimally processed fruit. Thus, this system could be Poynton, C. A. (1996). A technical introduction to digital video. John Wiley yy&
used to measure, in real-time, the color quality in foods and choose Sons, Inc.
1226 ~ a et al. / LWT - Food Science and Technology 59 (2014) 1220e1226
E. Saldan

Robards, K., Prenzler, P. D., Tucker, G., Swatsitang, P., & Glover, W. (1999). Phenolic Takenaka, M., Yan, X. J., Ono, H., Yoshida, M., Nagata, T., & Nakanishi, T. (2003).
compounds and their role in oxidative processes in fruits. Food Chemistry, 66(4), Caffeic acid derivatives in the roots of yacon (Smallanthus sonchifolius). Journal
401e436. http://dx.doi.org/10.1016/s0308-8146(99)00093-x. of Agricultural and Food Chemistry, 51(3), 793e796. http://dx.doi.org/10.1021/
Saldan~ a, E., Siche, R., Lujan, M., & Quevedo, R. (2013). Review: computer vision jf0207351.
applied to the inspection and quality control of fruits and vegetables. Brazil Yan, X. J., Suzuki, M., Ohnishi-Kameyama, M., Sada, Y., Nakanishi, T., & Nagata, T.
Journal of Food Technology, 16(4), 254e272. http://dx.doi.org/10.1590/S1981- (1999). Extraction and identification of antioxidants in the roots of yacon
67232013005000031. (Smallanthus sonchifolius). Journal of Agricultural and Food Chemistry, 47(11),
Sapers, G. M., & Douglas, F. W. (1987). Measurement of enzymatic browning at cut 4711e4713. http://dx.doi.org/10.1021/jf981305o.
surfaces and in juice of raw apple and pear fruits. Journal of Food Science, 52(5), Zou, X. B., Zhao, J. W., & Li, Y. X. (2007). Apple color grading based on organization
1258e1285. feature parameters. Pattern Recognition Letters, 28(15), 2046e2053. http://
Segnini, S., Dejmek, P., & Oste, R. (1999). A low cost video technique for colour dx.doi.org/10.1016/j.patrec.2007.06.001.
measurement of potato chips. Food Science and Technology-Lebensmittel-
Wissenschaft & Technologie, 32(4), 216e222.

You might also like