Journal of Food Engineering 78 (2007) 22–26

www.elsevier.com/locate/jfoodeng

Modeling of color values for nondestructive evaluation

of maturity of mango
a,*
S.N. Jha , Sangeeta Chopra a, A.R.P. Kingsly b

a
Agriculture Structure and Process Engineering Division, Central Institute of Postharvest Engineering and Technology,
PO PAU, Ludhiana 141 004, India
b
Horticulture crop processing division, CIPHET, Abohar 152 116, Punjab, India

Received 20 May 2005; accepted 24 August 2005

Available online 19 October 2005

Abstract

Mango, an important tropical fruit, is marketed throughout the world. Quality of ripened mango depends on its maturity at harvest,
which is hitherto determined subjectively by experience. In the present investigation, the potential of a nondestructive method for pre-
dicting maturity using color values taken by a handheld colorimeter was explored. Maturity of mango was defined in term of maturity
index (Im), a ratio of total soluble solids (TSS) of mango under experiments and minimum level of TSS (8 Brix) of matured mango,
expressed in percentage. Hunter L, a, and b values of 315 freshly harvested mangoes ranging from immature to over matured were mea-
sured using HunterLab colorimeter. TSS of whole mango juice was measured by a handheld refractometer and maturity index was com-
puted. The maturity index and L, a, and b values of 160 samples were fitted in different forms of models using multiple linear regression
(MLR), partial least square and principal component regression. The model using MLR on variables a, b and the product ab was
selected. The prediction performance of the model was tested with another set of 100 unripe samples. Precision of prediction was also
verified by sensory evaluation of 55 ripe mangoes and was found that the fruits predicted to be mature could ripe with high-satisfied taste
while the ones predicted to be immature or over mature were mostly rejected by the panels.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Maturity index; Total soluble solids; Color values; Regression models; Nondestructive method

1. Introduction Maturity has been correlated with various physical

Mango (Mangifera indica L.) is an important tropical
fruit having heavy demand in world market. Taste, physi-
cal appearance and postharvest shelf life of fruit are depen-
dent upon maturity level at harvest. When fruits are
harvested before maturity they do not ripen uniformly
and may present excessive shrinkage and low levels of
sweetness. Even ethylene or acetylene treatment cannot in-
duce complete ripening with proper aroma, flavour, and
taste in immature fruits. On the other hand fruits harvested
at late maturity stage, result in reduced shelf life with great-
er susceptibility to disease.
*
Corresponding author. Tel.: + 91 161 2572039.
E-mail address: snjha_ciphet@yahoo.co.in (S.N. Jha).
characteristics like skin color, shape, size and shoulder
growth. Attempts have been made to fix a maturity
standard for mango based on specific gravity (Tandon,
Kalra, & Singh, 1988) and firmness (Samson, 1980) but
inconsistency in these parameters of mango varieties
restricted its use as a criterion to predict maturity (Tandon
& Kalra, 1983, 1986). Some chemical parameters like
soluble solids content (TSS), titratable acidity, starch,
phenolic compounds, carotenoids and dry matter content
have also been used to determine maturity of mango
(Jha, Kingsly, & Chopra, 2006). TSS above 8% and acidity
about 1% are some indices for assessing mango maturity.
Many workers have worked for maturity and quality
indices of other fruits and vegetables too (Lee, 1981), but
most of them are of chemical or physiological nature,

0260-8774/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfoodeng.2005.08.048
 S.N. Jha et al. / Journal of Food Engineering 78 (2007) 22–26
and their determination involves very laborious laboratory
techniques.
Nowadays, nondestructive techniques for quality evalu-
ation have gained in popularity (Iwamoto, Kawano, &
Ozaki, 1995; Jha & Matsuoka, 2000, 2004). These tech-
niques, particularly for fruits and vegetables, are quick
and easy to use. Many physical characteristics of fruits
and vegetables have been determined nondestructively
(Kato, 1997; Lammertyn, Nicolai, Ooms, Smedt, &
Baerdemaeker, 1998; Lesage & Destain, 1996; Nussinovitch,
Ward, & Mey-tal, 1996). For mango some such efforts
using near-infrared (NIR) spectroscopy (Guthrie & Walsh,
1997; Saranwong, Sornsrivichai, & Kawano, 2001, 2003a,
2003b, 2004; Walsh, Golic, & Greensill, 2004), visual spec-
tral analysis (Jha, Chopra, & Kingsly, 2005; Schmilovitch,
Mizarach, Hoffman, Egozi, & Fuchs, 2003), acoustic and
ultrasound techniques (Mizrach, Flitsanov, Schmilovitch,
& Fuchs, 1999; Mizrach, 2000; Polderdijk, Kho, Kruif, &
Carmi, 2000; Shmulevich, Galili, & Howarth, 2003;
Valente & Ferrandis, 2003) are also reported, but these
instruments are costly and difficult to carry to the mango
orchards. A hand held colorimeter is cheaper and easier
to carry to the orchards. Medlicott, Semple, Thompson,
Black bourne, and Thompson (1992) have reported a sig-
nificant correlation between peel color score (visual assess-
ment) and Hunter a/b ratio for mango, which gives an idea
of maturity. A calibration equation related to maturity
level based on some chemicals at harvest and color values
must be developed for use with the portable colorimeter
to replace the typical subjective method and reduce the
problem of carrying the instruments to the field.
The objectives of this paper were thus to define a matu-
rity index based on mango TSS, develop the calibration
equation using color values and evaluate the calibration
efficiency by examining the eating quality of fruits to ex-
plore the potential of prediction of the index nondestruc-
tively using a robust and portable colorimeter.
2. Materials and methods
2.1. Sampling of mango
Three mango (cv. Dashehari) orchards from different
locations of the university farm were selected based on
flowering amount, size of trees (for easy harvest) and loca-
tion (easily approachable) for the study in spring 2003 to
2004. Fruiting of mango was continuously observed. Four
mangoes from different directions of selected trees, situated
at different location of the orchard, were randomly har-
vested at 48 h intervals from the 6th June in 2003 and the
27th May 2004 (when mango was not fully mature) to their
final harvesting dates (when the majority of the mangoes
was ripening on the tree) of respective years, and they were
brought to laboratory and kept at 35 ± 1 C and 65 ± 1%
relative humidity (RH) to equilibrate with environment be-
fore experimentation. Altogether 315 samples were col-
lected from the orchards for experiment.
23
2.2. Experimentation
The HunterLab color values of freshly harvested man-
goes were obtained in terms of L, a, and b values using a
HunterLab Mini Scan XE plus colorimeter (HAL, USA,
model 45%-L). In the Hunter scale, ÔLÕ measures lightness
and varies from 100 for perfect white to zero for black,
approximately as the eye would evaluate it. The chromatic-
ity dimensions ÔaÕ measures redness when positive, gray
when zero, and greenness when negative, and ÔbÕ measures
yellowness when positive, gray when zero, and blueness
when negative.
The colorimeter was calibrated with standard black and
white calibrations tiles provided with the instrument and
displayed values were matched with the values reported
in the operating manuals. The setup of the colorimeter
was changed to color scale and the nose cone was kept in
complete contact with the mango surface to prevent leak-
age of light emitted by the colorimeter. The color values
L, a, b were stored automatically in the instrument. The
experiment was repeated at four places; located near the
apex and the stem regions of each mango and average
values were used for further calculation. Immediately after
getting the color values, the mango was peeled (0.2 mm
thick) and squeezed with a domestic juicer at room temper-
ature (32.2 C). Suspended particles were removed from the
juice by filtering through the muslin cloth. The TSS value
was measured from the juice with a handheld refractometer
(ERMA, Japan). Average TSS value from duplicate mea-
surements were used for the computation of maturity index
(Im, %) as follows:
TSS
Im ¼  100 ð1Þ
8
The Im was correlated with L, a and b values to know
maturity level of mango before harvest. If it falls below
or above 100%, mangoes were considered to be under or
over matured, respectively.
A random portion (55 total number) of daily harvest of
known maturity was stored at 35 ± 1 C temperature and
65 ± 1% RH for a week for ripening; and a sensory evalu-
ation of ripe mangoes by a trained panel of the institute
using a 9 point hedonic scale was conducted to verify the
hypothesis that the mangoes having the computed maturity
of about 100% are actually matured and ripened properly.
Mature mangoes ripen properly and are liked more as com-
pared to relatively immature and over matured ones, which
do not ripe properly.
2.3. Data analysis
Acquired L, a, and b values were imported to MS-excel
software from the instrument and then to the Unscrambler
(CAMO AS, Trondheim, Norway, version 8.0.5), a statis-
tical software package, for multivariate calibration. The
Im was calculated from TSS values for all 315 samples
and they were split randomly into three sets (Table 1):
24 S.N. Jha et al. / Journal of Food Engineering 78 (2007) 22–26

Table 1 based on their degree of likeness, to ascertain the relation-

Statistical characteristics of maturity index (Im) of samples used in data ship between the predicted maturity index and the eating
analysis
quality of ripe mango were computed.
Sample sets No. of Range of Im, % Mean Im, % Standard
samples deviation, %
3. Results and discussion
Calibration 160 75.0–237 129.9 46.0
Validation 100 72.5–225 112.0 38.7
Sensory test 55 70.0–230 118.0 40.6
3.1. Calibration

The standard error of calibration (SEC) and standard

calibration set of 160 samples of Im ranging from 75% to
237%, mean 129.9% and standard deviation 46%; valida-
tion set of 100 samples of Im ranging from 72.5% to
225%, mean 112% and standard deviation 38.7% and a
set of 55 samples of Im ranging from 70% to 230% for con-
ducting sensory evaluation after ripening at known temper-
ature and RH for a week before further analysis of data.
In order to search for the best relationship for predicting
Im from L, a, and b color values, six models (Table 2) were
fitted to the data using multiple linear regressions (MLR).
The best among them, based on standard error of calibra-
tion (SEC), validation (SEP), multiple correlation coeffi-
cients R and convenience in terms of computation on
farm, was selected. The performance of the calibration
models for prediction of Im was tested with validation set
of samples. Variables of selected model were then analyzed
using partial least square regression (PLS), principal com-
ponent regression (PCR) techniques to check the better
predictability. Sensory scores for ripe mangoes were
grouped for different ranges, at 10% interval, of maturity
index at harvest. Percentages of respondents in each group,
Table 2
Generalized forms of Im models tested in terms of color values (L, a, b)
Model Variables Models*
No.
1 a, b, L Im = C1 + C2a + C3b + C4L
2 a, b I m = C1 + C2 a + C3 b
3 a, b, a · b Im = C1 + C2a + C3b + C4ab
4 a, b, a2, b2, a · b Im = C1 + C2a + C3b + C4ab + C5a2 + C6b2
5 a2, b2 Im = C1 + C2a2 + C3b2
6 b, a · b Im = C1 + C2b + C3ab
*
C1, C2, C3 are modelsÕ constants.
error of prediction (SEP) values were found to be the low-
est and the multiple correlation coefficient though highest
for the MLR model number 4 (Table 3), it involves higher
order of variables and thus may be difficult to compute in-
stantly at farm level. The next model having very close or
similar SEC/SEP and R values was the model number 3,
besides simpler in nature for computation. Thus, the model
number 3 was selected and tested further for calibration
using PLS and PCR techniques. PLS and PCR model using
same variables could not produce better performance as
compared to earlier selected MLR model (Table 4).
Normally PCR and PLS are effective for reducing the
dimension of data that contains large number of variables.
The model number 3 actually has only 3 variables and thus
PCR and PLS did not provide any improvement.
The best model using MLR on variables, a, b, and prod-
uct of a and b was selected for predicting Im. It indicates
that Im is dependent only on a, b and interaction of a
and b color values. Negative values of ÔaÕ only measures
greenness while ÔbÕ shows yellowness when positive and
hence these values (green to yellow color) play more signif-
icant role in determining maturity index of mango. Singh
(1973) has also observed that the change in peel color from
green to yellow could be correlated with maturity. L, a
measure of brightness, i.e. from black to white was not
so significant in predicting Im as can be seen from poor per-
formance of model number 1 having L term (Table 3).
3.2. Validation and relationship between predicted maturity
and eating quality
To check the performance of the selected calibration
model number 3, prediction results using validation set of
100 samples were compared. The calibration and valida-
tion results show that the SEC, SEP and correlation coef-

Table 3
Statistical results of various MLR models for determining maturity index of mango
Model No. SEC/SEP* Bias Correlation
Calibration Validation Calibration Validation Calibration Validation
1 13.60 12.19 0.00 0.13 0.9 0.89
2 13.93 12.53 0.00 0.33 0.89 0.88
3 13.15 10.72 0.00 0.26 0.91 0.91
4 12.70 10.72 0.00 0.11 0.91 0.92
5 15.76 15.52 0.00 0.62 0.86 0.81
6 13.39 11.59 0.00 0.29 0.90 0.90
*
No of samples used in calibration and validation were 160 and 100, respectively. Ranges of maturity index in calibration and validation were 75–237%
and 72.5–225%, respectively.
 S.N. Jha et al. / Journal of Food Engineering 78 (2007) 22–26 25

Table 4
Effect of regression methods on performance of selected model
Variables Regression methods SEC/SEP* Correlation Bias
Calibration Validation Calibration Validation Calibration Validation
a, b, a · b MLR 13.15 10.72 0.91 0.92 0.00 0.26
PLS 13.47 11.72 0.90 0.90 0.00 0.30
PCR 13.47 11.73 0.90 0.90 0.00 0.30
*
No of samples used in calibration and validation were 160 and 100, respectively. Ranges of maturity index in calibration and validation were 75–237%
and 72.5–225%, respectively.

250
250
Predicted maturity index, %

Predicted maturity index, %

200
200

150
150

100
100

50 50
50 100 150 200 250 50 100 150 200 250
Actual maturity index, % Actual maturity index, %
Fig. 1. Actual versus predicted maturity index of calibration sample set.
Fig. 2. Actual versus predicted maturity index of mango of validation
sample set.

ficients R are similar with no significant bias between them

(Table 4). A comparison of scatter plots of actual Im values
versus computed one for calibration (Fig. 1) and validation
(Fig. 2) sets of samples also showed that the color values
and the destructive standard methods gave very close
results.
In literature no model for predicting maturity index
using color values of mango is available to compare with
the present results. The scores of sensory evaluation, how-
ever, confirms that all samples having predicted maturity
index between 91% and 110% were either in categories of
‘‘liked very much’’ or ‘‘liked’’ by the panel members with
having comments of excellent taste (Table 5). Relatively
immature samples did not ripe properly and thus 41% of
them fell under category of ‘‘neither liked nor disliked’’
and 1% in the ‘‘disliked’’ category. The Table 5 reveals that
the percentage of rotten and disliked samples increased
with increase in predicted maturity index, which is in agree-
ment with the reported literature that immature mango
does not ripen properly while over matured one ripens fas-
ter and is having low shelf life (Kalra & Tandon, 1983).
This also confirms the hypothesis of computation of matu-
rity index and prediction of the same by the selected model
based on color values acquired by the colorimeter. The
developed MLR model using color values a, b, and a · b,
thus may be able to determine the maturity index of intact
mango at harvest.

Table 5
Sensory scores for ripe mangoes of known maturity index (Im) estimated at harvest
Predicted Im, % Percentage of respondents of sensory panel
Liked very much* Liked Neither liked nor disliked Disliked Comments
70–90 34 24 41 1 Improper ripening
91–110 69 31 0 0 Excellent taste
111–130 44 27 9 20 A little over- ripening
131–150 25 20 14 31 Over-ripened
151–170 15 14 10 45 16% samples rotted
171–190 30 5 8 54 30% samples rotted
191–210 0 5 0 59 41% samples rotted
211–230 0 0 0 52 48% samples rotted
*
MangoesÕ ripening conditions: Room temperature 35 ± 1 C and relative humidity 65 ± 1%.
26 S.N. Jha et al. / Journal of Food Engineering 78 (2007) 22–26
4. Conclusions
For practical and low cost application, maturity of
mangoes was predicted from peel color measured at har-
vesting date. The maturity index calculated from the ratio
of flesh Brix to ideal Brix of mature green fruit (8 Brix)
was correlated with color values and their products. By
multiple linear regression (MLR), precise calibration equa-
tion could be obtained. Precision of the developed equa-
tion was proved by the sensory evaluation of ripe
mangoes. It was found that the fruits predicted to be ma-
ture could ripe with high-satisfied taste while the ones pre-
dicted to be immature were mostly rejected by the panels.
For commercial use however addition of more number of
samples and varieties to the model is desirable to exploit
the full potential of colorimeter for prediction of maturity
of mango nondestructively.
References
Guthrie, J., & Walsh, K. (1997). Non-invasive assessment of pineapple
and mango fruit quality using near-infrared spectroscopy. Australian
Journal of Experimental Agriculture, 37, 253–263.
Iwamoto, M., Kawano, S., & Ozaki, Y. (1995). An overview of research
and development of near infrared spectroscopy in Japan. Journal of
Near Infrared Spectroscopy, 3, 179–189.
Jha, S. N., & Matsuoka, T. (2004). Nondestructive determination of acid
brix ratio of tomato juice using near infrared spectroscopy. Interna-
tional Journal of Food Science and Technology, 39, 425–430.
Jha, S. N., Chopra, S., & Kingsly, A. R. P. (2005). Determination of
sweetness of intact mango using visual spectral analysis. Biosystems
Engineering, 91, 157–161.
Jha, S. N., Kingsly, A. R. P., & Chopra, S. (2006). Physical and
mechanical properties of mango during growth and storage for
determination of maturity. Journal of Food Engineering, 72, 73–76.
Jha, S. N., & Matsuoka, T. (2000). Non-destructive techniques for quality
evaluation of intact fruits and vegetables. Food Science and Technology
Research, 6(4), 248–251.
Kalra, S. K., & Tandon, D. K. (1983). Ripening behaviour of Dashehari
mango in relation to harvest period. Horticulture Science, 19, 263.
Kato, K. (1997). Electrical density sorting and estimation of soluble solids
content of watermelon. Journal of Agricultural Engineering Research,
67, 161–170.
Lammertyn, J., Nicolai, B., Ooms, K., Smedt, V. de., & Baerdemaeker, J.
de. (1998). Nondestructive measurement of acidity, soluble solids, and
firmness of jonagold apples using NIR-spectroscopy. Transactions of
ASAE, 41(4), 1089–1094.
Lee, S. K. (1981). A review and background of the avocado maturity
standard. Yearbook California Avocado Society, 65, 101–109.
Lesage, P., & Destain, M. F. (1996). Measurement of tomato firmness by
using a non-destructive mechanical sensor. Postharvest Biology and
Technology, 8, 45–55.
Medlicott, A. P., Semple, J., Thompson, A. J., Black bourne, H. R., &
Thompson, A. K. (1992). Measurement of color changes in ripening
bananas and mangoes by instrumental, chemical and visual assess-
ments. Tropical Agriculture (Trinidad and Tobago), 69, 161–166.
Mizrach, A., Flitsanov, U., Schmilovitch, Z., & Fuchs, Y. (1999).
Determination of mango physiological indices by mechanical wave
analysis. Postharvest Biology and Technology, 16, 179–186.
Mizrach, A. (2000). Determination of avocado and mango fruit properties
by ultrasonic technique. Ultrasonics, 38, 717–722.
Nussinovitch, A., Ward, G., & Mey-tal, E. (1996). Gloss of fruits
and vegetables. Lebensmittel-Wissenschaft und Technologie, 29, 184–
186.
Polderdijk, J. J., Kho, R. M., Kruif, A. P. M., & Carmi, Y. (2000).
Firmness of mangoes (Mangifera indica L) measured acoustically,
mechanically and manually. Acta Horticulturae, Wageningen, 509,
861–865.
Samson, J. A. (1980). Tropical fruits. London: Longman.
Saranwong, S., Sornsrivichai, J., & Kawano, S. (2001). Improvement of
PLS calibration for brix value and dry matter of mango using
information from MLR calibration. Journal of Near-Infrared Spec-
troscopy, 9, 287–295.
Saranwong, S., Sornsrivichai, J., & Kawano, S. (2003a). On tree
evaluation of harvesting quality of mango fruit using a hand held
NIR instrument. Journal of Near-Infrared Spectroscopy, 11, 283–
293.
Saranwong, S., Sornsrivichai, J., & Kawano, S. (2003b). Performance of
portable near-infrared instrument for brix value determination of
intact mango fruit. Journal of Near-Infrared Spectroscopy, 11, 175–
181.
Saranwong, S., Sornsrivichai, J., & Kawano, S. (2004). Prediction of ripe
stage eating quality of mango fruit from its harvest quality measured
nondestructively by near-infrared spectroscopy. Postharvest Biology
and Technology, 31, 137–145.
Schmilovitch, ZeÕev., Mizarach, A., Hoffman, A., Egozi, H., & Fuchs, Y.
(2003). Determination of mango physiological indices by near-infrared
spectrometry. Postharvest Biology and Technology, 19, 245–252.
Shmulevich, I., Galili, N., & Howarth, M. S. (2003). Nondestructive
dynamic testing of apples for firmness evaluation. Postharvest Biology
and Technology, 29, 287–299.
Singh, R. N. (1973). Mango. New Delhi: Indian Council of Agricultural
Research.
Tandon, D. K., & Kalra, S. K. (1983). Studies on maturity of developing
mango fruit cv. Dashehari. Plant Physiology Biochemistry, 10, 113.
Tandon, D. K., & Kalra, S. K. (1986). Studies on developing mango fruits
to assess maturity. Indian Journal of Horticulture, 43, 51.
Tandon, D. K., Kalra, S. K., & Singh, B. P. (1988). Ripening pattern of
specific gravity graded Dashehari mangoes. Indian Journal of Horti-
culture, 45(3), 219.
Valente, M., & Ferrandis, J. Y. (2003). Evaluation of textural properties of
mango tissue by near-field acoustic method. Postharvest Biology and
Technology, 29, 219–228.
Walsh, K. B., Golic, M., & Greensill, C. V. (2004). Sorting of fruit using
near infrared spectroscopy: application to a range of fruit and
vegetables for soluble solids and dry matter content. Journal of Near
Infrared Spectroscopy, 12, 141–148.