with an optical fibre in interactance mode. Slaughter 1995 determined that visible and NIR-
spectroscopy could be used to measure non-de- structively the internal quality of peaches and
nectarines as characterised by their soluble solid SSC, sugar, sorbitol and chlorophyll contents.
Bellon-Maurel 1992 used the wavelength region between 800 – 1050 nm to built a model for sugar
measurement. Recently, Moons et al. 1997 and Lammertyn et al. 1998 established a relationship
between NIR spectra and apple fruit quality parameters such as acidity, pH, sugar content and
texture parameters.
For internal quality measurements, it is impor- tant that the NIR radiation penetrates the apple
tissue sufficiently, an issue not often discussed in the literature. Chen and Nattuvetty 1980 investi-
gated the effect of the distance between the light incident and detection points on the transmittance
and on the depth through which the detected light penetrated into the fruit. Hother et al. 1995
followed the changes in reflectance properties of unpeeled apple disks of varying thickness. They
found that, depending on the variety and the wavelength, the penetration depth varied between
0 and 7 mm. For ‘Jonagold’ apples, the maximum depth was 5.5 mm. However, the authors only
considered a wavelength range between 480 and 800 nm. It should also be noted that, even if the
radiation sufficiently penetrates the apple tissue, reflected radiation due to internal scattering needs
to be separated from that due to specular reflec- tion. Several optical configurations have been
used in the literature, including bifurcated light guides and 0°45° configurations. It is not clear
how the configuration affects the quality of the calibration models.
The objectives of this paper were i to compare two optical configurations for measuring internal
apple quality attributes by means of NIR reflec- tance measurements, ii to obtain fruit skin
parameters which describe the interaction of the skin and the incident radiation, and iii to deter-
mine the penetration depth values of NIR radia- tion in apple tissue for the wavelengths between
500 and 1900 nm.
2. Materials and methods
2
.
1
. Fruit Apples Malus domestica Borkh. cv. Elstar
used for the first experiment were purchased at a local auction and stored for 2 days at 20°C and
70 relative humidity to equilibrate. For the cali- bration models, 60 apples were used. From each
apple, four spectra 880 – 1650 nm with the bifur- cated and with the 0°45°-configuration see fur-
ther were taken at exactly the same position. The soluble solids content, which is strongly correlated
with the sugar content, was measured at the same positions with a digital refractometer PR-101
Palette Series, ATAGO CO., Ltd., Japan.
‘Jonagold’ apples, used for the light penetration experiments, were also purchased at a local auc-
tion and measured after 2 days of equilibration at 20°C and 70 relative humidity.
2
.
2
. Reflectance measurements For each apple, four reflection spectra 880 –
1650 nm, wavelength increment 0.5 nm were taken at four equidistant positions around the
equator, with a spectrophotometer Optical Spec- trum Analyser OSA 6602, Rees Instruments
Ltd., Godalming, UK. The light source consisted of a 12 V100 W tungsten halogen lamp Philips
7724.M28 that could be used both in the visible and near infrared region. Two different optical
configurations were used. With the bifurcated op- tical configuration type MIO-6134 the light is
guided to the sample by source fibres, and from the sample with the detector fibres. In the head of
the bifurcated cable, the source and detector fibres are situated randomly Fig. 1A. The fibre has
an active surface of 4 mm
2
and was held directly on the skin of the fruit. This has the advantage of
a higher light intensity than a non-contact source. It can only be used in the wavelength range from
380 to 1650 nm. The 0°45° optical configuration Fig. 1B consists of a black box type 6151 in
which the source and the detector fibres are posi- tioned at an angle of 45°. The incident beam falls
perpendicularly onto the sample and is detected under an angle of 45°, to avoid specular reflec-
tion. Since the illuminated surface is larger, the intensity will be lower than with the bifurcated
cable for a given light intensity. In both cases the reflected light is divided into individual wave-
lengths
by the
diffracting gratings
of the
monochromator. Grating A is used for the wave- length range from 300 to 1080 nm and grating B
for the range 1080 – 2000 nm. A silicon detector was used for the visible and the beginning of the
near infrared range 300 – 1100 nm and a PbS detector was used in the NIR range 1000 – 2000
nm. To compare both optical configurations the 0°45° device was only used in the 880 – 1650 nm
range. The signals were processed with software, model 6857 v1.30. The configuration was cali-
brated with a HeNe laser and a spectrum from a BaSO
4
-disc served as reference.
2
.
3
. Measurement of internal apple quality An average spectrum was calculated for each
apple. The average spectra were pre-processed by reducing the number of points of measurement
and taking the second derivative using the method of Savitzky – Golay Savitzky and Golay, 1964.
The second derivative or multiplicative scatter correction MSC was used to correct for additive
and multiplicative effects in the spectra Martens and Naes, 1989. The technique used for the
calibration was partial least squares PLS Haa- land and Thomas, 1988. The calculations were
carried out using ‘The Unscrambler’ CAMO, AS, Trondheim, Norway, a statistical software pack-
age for multivariate calibration.
2
.
4
. Skin reflectance and transmission properties To obtain information about the light penetra-
tion properties, an experiment based on earlier work of Lillesaeter 1982 was performed. Lille-
saeter 1982 divided the information in a reflec- tance spectrum of a leaf into two components:
information coming from the leaf and informa- tion coming from the background. In the present
study the leaf surface was replaced by the skin of the apple and the background corresponded to
the tissue under the skin.
The total reflected radiation R
tot
consists of two components: the skin component, R
skin
, which is the radiation reflected by the skin with a
perfectly black background, and a background component, being the radiation reflected by a
non-black background, changed by transmission through the skin. With the intensity of the inci-
dent light beam and t, the transmission of the skin, the following formula can be derived:
R
tot
= R
skin
+ R
back
= r
skin
I + r
back
t
2
I 1
with r
skin
and r
back
the reflectance of the skin and the background, respectively. The transmission
parameter t is squared because the incident light crosses the skin twice before it is detected. Eq. 1
is a simplification of the formula for total reflec- tion from a thin layer Kortu¨m, 1969. The de-
nominator of the second term 1 − r
skin
r
back
is omitted here. This can be done when r
skin
r
back
1.
Fig. 1. The bifurcated optical configuration A and the 0°45° optical configuration B.
Table 1 Survey of the prediction performance of the different calibration SSC models for the bifurcated and the 0°45° optical
configurations
a
Lat. Var. Model
RMSEC Pre-treatment
RMSEP Correlation
Bifurcated optical configuration 1
5 10
0.45 0.55
0.91 c
5 7
0.37 0.73
0.85 2
10 c
10 7
10 0.43
3 0.59
0.90 10
4 c
15 6
0.43 0.57
0.91 5
c 20
10 7
0.47 0.62
0.88 °
45
° optical configuration 6
5 10
0.61 0.72
0.83 c
5 7
10 7
0.39 0.84
0.79 c
10 7
10 0.52
8 0.70
0.86 10
9 c
15 6
0.53 0.65
0.87 10
c 20
10 5
0.56 0.66
0.87
a
x indicates the size x of the reduction of the original spectrum; c x denotes the half of the interval x used for the calculation of the second derivative using the method of Savitzky–Golay and ‘’ indicates that multiplicative scatter correction MSC has been
applied.
The total reflectance is defined as: r
tot
= R
tot
I =
r
skin
+ r
back
t
2
2 Two measurements with a different background
back1 and back2 are sufficient to solve the equa- tions in the skin parameters t and r
skin
. r
tot,back1
= r
skin
+ r
back1
t
2
3 r
tot,back2
= r
skin
+ r
back2
t
2
4 t =
r
tot,back1
− r
tot,back2
r
back1
− r
back2
5 r
skin
= r
tot,back1
− r
back1
t
2
6 All the measurements for this test were exe-
cuted with the spectrophotometer as described above using the 0°45° optical configuration. For
this test a piece of the red and the green side of the apple were used. The skin was carefully iso-
lated from the fruit flesh with a razor blade and used for the measurements with the different
backgrounds.
2
.
5
. Penetration depth as a function of the wa6elength
In a third experiment, the light penetration depth in apple tissue was evaluated for each wave-
length l in the range from 500 to 1900 nm. The test was performed on ‘Jonagold’ apples. The
apple was cut in two and on the green skin side, ten spectra were taken in the wavelength range
from 500 to 1900 nm with the 0°45° optical configuration. During the measurements, the
slices were protected from drying air movement and external light by a black box cover. Subse-
quently, a thin slice of apple tissue was removed with a professional slice cutter Graef, Germany
and again ten spectra were taken on the skin side of the apple. The thickness of the slice was mea-
sured with a calliper. This procedure was repeated several times until all the tissue was removed from
the skin. Finally, ten spectra of the skin were taken. The different apple slice thickness values
u were 3.5, 2.78, 2.48, 2.2, 2.02, 1.68, 1.47, 1.25, 1.15, 1.02, 0.93, 0.82, 0.73, 0.59, 0.43, 0.3, 0.18,
0.1 and 0.04 cm. An average spectrum was calcu-
lated for each thickness. For each wavelength and thickness the standard deviation was calculated
based on the ten replications.
3. Results and discussion
