sensors
Article
A Novel Method for Calibration of Digital Soil Images
Captured under Irregular Lighting Conditions
Sung-Ha Baek 1 , Ka-Hyun Park 2 , Jun-Seo Jeon 2 and Tae-Young Kwak 2, *
Citation: Baek, S.-H.; Park, K.-H.;
Jeon, J.-S.; Kwak, T.-Y. A Novel
Method for Calibration of Digital Soil
Images Captured under Irregular
Lighting Conditions. Sensors 2023, 23,
296. https://doi.org/10.3390/
s23010296
Academic Editor: Benoit Vozel
Received: 18 October 2022
Revised: 9 December 2022
Accepted: 24 December 2022
Published: 27 December 2022
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Sensors 2023, 23, 296. https://doi.org/10.3390/s23010296
1 School of Civil and Environmental Engineering & Construction Engineering Research Institute,
Hankyong National University, Anseong 17579, Republic of Korea
2 Department of Geotechnical Engineering Research, Korea Institute of Civil Engineering and Building
Technology, Goyang 10223, Republic of Korea
* Correspondence: tykwak@kict.re.kr; Tel.: +82-10-3922-5661
Abstract: Soil color is commonly used as an indicator to classify soil and identify its properties.
However, color-based soil assessments are susceptible to variations in light conditions and the
subjectivity of visual evaluations. This study proposes a novel method of calibrating digital images
of soil, regardless of lighting conditions, to ensure accurate identification. Two different color space
models, RGB and CIELAB, were assessed in terms of their potential utility in calibrating changes to
soil color in digital images. The latter system was determined to be suitable, as a result of its ability to
accurately reflect illuminance and color temperature. Linear regression equations relating soil color
and light conditions were developed based on digital images of four different types of soil samples,
each photographed under 15 different light conditions. The proposed method can be applied to
calibrate variations in the soil color obtained by digital images, thus allowing for more standardized,
objective, and accurate classification and evaluation of soil based on its color.
Keywords: soil color; CIELAB color model; RGB color model; illuminance; color temperature
1. Introduction
Because the color of soil varies depending on the constituent minerals, organic matter
content, water content, and ion concentration, it has been widely used as a basic indicator
to classify soil and predict its properties [1,2]. For example, in the field of agriculture,
the soil color is used as a representative soil classification index [3], and different types
of crops and farming types are monitored based on the soil color. In addition, in civil
engineering, because similarly colored soils in adjacent areas are highly likely to have
similar characteristics, the colors of the soil samples collected during ground surveys are
recorded on the corresponding drilling logs.
Soil color is determined through observation with the naked eye. The Munsell soil color
chart was developed to objectively distinguish soil colors observed using human vision, and
represents colors as a combination of hue, value, and chroma [4]. ASTM [5] requires that the
soil color be determined (indicated in the order of color brightness/saturation) by finding the
standard color chip most similar to the color of the soil sample in the Munsell soil color book,
which is the most widely used soil color determination method to date [6–9]. However,
determining soil color using the Munsell soil color book has the following shortcomings:
1. When finding the standard color chip most similar to the color of the soil sample with
the naked eye, the results are highly likely to be affected by the subjective perception
or visual sensitivity of the observer, and the process is time-intensive [10];
2. Because the color of the soil sample and the standard color chip vary depending on
environmental factors, such as illumination conditions, it is difficult to determine the
absolute color [10,11];
3. Because the standard color chip of the Munsell book is divided into discrete elements,
numerical and statistical analysis are difficult [12].
https://www.mdpi.com/journal/sensors
Sensors 2023, 23, 296 2 of 16

Digital image processing has recently received attention as a method to overcome this
limitation [13–15]. Digital image processing entails a series of processes to obtain desired
information by analyzing a digital image. A digital image consists of RGB (red, green, and
blue) values from 0 to 255 assigned to each pixel (in the case of a black-and-white image,
the pixels are instead assigned gray values from 0 to 255). Because digital image processing
is a computerized process, it is possible to quickly and objectively determine soil color
without the intervention of an observer, and because the color is displayed as a continuous
value, numerical and statistical analysis are possible.
Accordingly, many researchers are conducting research to obtain soil color using
digital image processing and to analyze the correlation between soil color and soil property.
Persson [16], Zanetti et al. [17], Santos et al. [18], Park [19], and Kim [20] reported the RGB
color intensity of soil obtained from digital images taken in an indoor studio and analyzed
the water content (or moisture content), and Zhu et al. [21] analyzed the correlation between
the gray color intensity and water content of soil obtained from black-and-white images.
These studies revealed that the RGB color intensity and gray color intensity tend to decrease
as the water content of the soil increases, and presented empirical equations capable of
predicting the water content from the soil color. In addition, Gomez-Robledo et al. [22]
and Moonrungsee et al. [23] obtained the RGB color intensity of soil from digital images
taken with a smartphone camera. From those results, the soil color was matched to that of
the Munsell soil color book [22] and was used to predict the content of phosphorus in the
soil [23]. However, although the above studies were able to rapidly identify soil color based
on digital image processing and successfully predict soil properties through this process,
they had the limitation of using images taken in an indoor studio under constant light
conditions. According to the principle of color expression, if the incident light changes, the
soil color changes, so it is difficult to apply these research results to images taken under
different light conditions, such as images taken at actual sites where it is impossible to
control the light conditions.
In this study, a new method was explored to calibrate the changes in apparent soil color
that occur according to changes in light conditions. An indoor studio that could simulate
the characteristics of natural light (illuminance and color temperature) was established,
and digital images were taken for four soil samples while changing the light conditions
15 times each (for a total of 60 shots). Digital image processing was performed to extract
the soil colors of the photographed samples in two different color space models (RGB,
CIELAB). We identified a color space model suitable for calibrating changes in soil color
according to light conditions, and finally developed a new method for calibrating soil color
in photographs taken under irregular light conditions.
The remainder of this paper is structured as follows. Section 2 describes our method-
ology, including the color space model, test equipment, digital imaging conditions, and
soil specimens. Section 3 describes the analysis results, followed by a description of the
proposed method. A conclusion is provided in Section 4.

2. Materials and Methods

2.1. Color Space Model for Digital Soil Images
Color is a three-dimensional psychophysical phenomenon that appears as light is
incident on an object and reflected. In general, “light” refers to visible light, which is an
electromagnetic wave with a wavelength between approximately 380 nm and 780 nm that
can be seen with the naked eye. When light is incident on an object (incident light), light of
some wavelengths is absorbed and light of other wavelengths is reflected (reflected light).
The color is determined according to the wavelength of the reflected light. For example,
plant leaves appear green because chlorophyll reflects electromagnetic waves in the green
wavelength band (approximately 500 to 570 nm), and the soil color appears red because the
soil particles reflect light in the red wavelength band (approximately 620 to 780 nm).
A method of numerically displaying the color of an object is called a color space
model. A color space model represents a specific color as a point on a one-dimensional
 because the soil particles reflect light in the red wavelength band (approximately 62
780 nm).
A method of numerically displaying the color of an object is called a color sp
model. A color space model represents a specific color as a point on a one-dimensi
Sensors 2023, 23, 296 axis or in a three-dimensional space, and the method of defining a one-dimensional 3 of 16

or three-dimensional space depends on the color space model (e.g., RGB, CIE XYZ,
CIELAB). Detailed definitions and theories of various color space models can be foun
axis or in a three-dimensional
Billmeyer and Saltzman [24], space, and the method
Wyszecki and Stilesof defining a one-dimensional
[25], Rossel axis chapter
et al. [26]. This
or three-dimensional space depends on the color space model (e.g., RGB, CIE XYZ, and
scribes the two color space models (RGB and CIELAB) used in this study for soil c
CIELAB). Detailed definitions and theories of various color space models can be found
analysis.
in Billmeyer and Saltzman [24], Wyszecki and Stiles [25], Rossel et al. [26]. This chap-
The RGB
ter describes the color system
two color spaceismodels
the most
(RGBwidely used approach
and CIELAB) used in thisinstudy
electronic
for soilequipm
such as digital cameras, and expresses colors using the three primary colors of light:
color analysis.
The RGB
(R), green color
(G), andsystem
blueis(B).
the most widely
In other used approach
words, each colorin electronic equipment,
is expressed as a such
combinatio
as digital cameras, and expresses colors using the three primary colors of light: red (R),
red, green, and blue. When more red, green, and blue are applied, the resulting colo
green (G), and blue (B). In other words, each color is expressed as a combination of red,
brighter; this is also called additive mixing. In the 8-bit digital system used by most e
green, and blue. When more red, green, and blue are applied, the resulting color is brighter;
tronic equipment, the RGB
this is also called additive colorIn
mixing. intensity
the 8-bitranges from 0 used
digital system (dark)
byto 255electronic
most (bright), and a t
of 16,777,216
equipment, the (=256 3) colors
RGB color can ranges
intensity be expressed [25]. to
from 0 (dark) Figure 1 shows
255 (bright), andthe RGBofcolor sp
a total
3 ) colors can be expressed [25]. Figure 1 shows the RGB color space model.
model. In a three-dimensional RGB color space, all colors correspond to a point with
16,777,216 (=256
In a three-dimensional
cube whose vertices are RGB color
red space,
(255, 0, 0),allgreen
colors(0,
correspond to a(0,
255, 0), blue point within
0, 255), a cube
black (0, 0, 0), w
whose vertices are red (255, 0, 0), green (0, 255, 0), blue (0, 0, 255), black (0, 0, 0), white (255,
(255, 255, 255), cyan (0, 255, 255), magenta (255, 0, 255), yellow (255, 255, 0).
255, 255), cyan (0, 255, 255), magenta (255, 0, 255), yellow (255, 255, 0).

Blue (B)
Magenta
Blue (255, 0, 255)
(0, 0, 255)

Cyan
(0, 255, 255) White
(255, 255, 255)

Red
(0, 0, 0) Red (R)
Black (255, 0, 0)

Green
(255, 255, 0)
(0, 255, 0)
Yellow
Green (G)
Figure 1. Three-dimensional RGB color space.
Figure 1. Three-dimensional RGB color space.
The RGB color space model has the advantage of being able to reproduce most colors
The RGB
in a simple way.color space model
In addition, has the
as mentioned advantage
above, becauseof beingelectronic
general able to reproduce
equipmentmost co
(e.g., digital cameras, smartphones) used to acquire digital images adopts
in a simple way. In addition, as mentioned above, because general electronic the RGB color equipm
space model, most existing studies [16–20,22,23] have obtained soil
(e.g., digital cameras, smartphones) used to acquire digital images adopts colors based on the
the RGB c
RGB color space model.
space model, most existing studies [16–20,22,23] have obtained soil colors based on
However, the RGB color space model is not capable of representing all human-
RGB color space
recognizable colors.model.
In addition, there is a problem in that the wavelength indicated by
However, the RGB color
these RGB values are different fromspace model isrecognized
the wavelengths not capable of cone
by the representing
cells (shortall
(S),human-
ognizable
medium (M), colors. In addition,
and long there isresponsible
(L) wavelengths) a problemforin human
that thevision.
wavelength indicated
Accordingly, in by th
1931, the Commission Internationale de l’Eclairage (CIE) established the
RGB values are different from the wavelengths recognized by the cone cells (short CIEXYZ color
space model, which can express all colors recognizable by humans based on the tristimulus
medium (M), and long (L) wavelengths) responsible for human vision. Accordingly
values recognized by human cone cells [27]. However, it is difficult to use the CIEXYZ
1931, the Commission
color space Internationale
model in engineering de(numerical
calculations l’Eclairage
and(CIE) established
statistical analysis ofthe CIEXYZ c
colors)
owing to perceptual non-linearity in the method of expressing the distance between colors
recognized by humans [25,26].
 lus values recognized by human cone cells [27]. However, it is difficult to use the CIE
color space model in engineering calculations (numerical and statistical analysis of co
owing to perceptual non-linearity in the method of expressing the distance between co
recognized by humans [25,26].
Sensors 2023, 23, 296 4 of 16
To overcome the above problem, CIE [28] proposed the CIELAB color space mo
based on the CIEXYZ color space model. Unlike the CIEXYZ color space model,
CIELAB color space model is an almost perceptually uniform color space model [26
To overcome the above problem, CIE [28] proposed the CIELAB color space model,
the CIELAB
based color space
on the CIEXYZ colormodel, colors Unlike
space model. are expressed
the CIEXYZ as combinations
color space model, of L*,the
a*, and b
CIELAB color space model is an almost perceptually uniform color space model [26]. In a* an
represents the brightness of the color, expressed from 0 (dark) to 100 (bright).
are CIELAB
the values representing the color,
color space model, colorswhere a* indicates
are expressed colors closer
as combinations toa*,
of L*, red
and(positive)
b*. L* or g
represents the brightness of the color, expressed from 0 (dark) to 100 (bright).
(negative), and b* indicates colors closer to yellow (positive) or blue (negative). Figu a* and b*
are values
shows therepresenting the color, where
three-dimensional CIELAB a* indicates colors closer
color space, to red
in which (positive)
every colororisgreen
represente
(negative), and b* indicates colors closer to yellow (positive) or blue (negative). Figure 2
a point. Although the CIELAB color space model has not been widely applied to di
shows the three-dimensional CIELAB color space, in which every color is represented by
aimage
point. analysis
Althoughcompared
the CIELABtocolorthe space
RGB model
color has
spacenotmodel, the CIELAB
been widely applied tocolor
digitalspace m
can express colors by dividing their components into “lightness”
image analysis compared to the RGB color space model, the CIELAB color space model can (L*-axis) and “chro
ticity” (a*-b*
express colors plane). Therefore,
by dividing it is expected
their components to be useful
into “lightness” in correcting
(L*-axis) changes in soil c
and “chromaticity”
(a*-b*
resulting from light conditions (illuminance indicating the lightnesssoil
plane). Therefore, it is expected to be useful in correcting changes in of color
light and c
resulting from light conditions (illuminance
temperature indicating the chromaticity of light). indicating the lightness of light and color
temperature indicating the chromaticity of light).

Figure2.2.Three-dimensional
Figure Three-dimensional CIELAB
CIELAB color space.
color space.

In this study, to convert the digital image acquired based on the RGB color space
In this study, to convert the digital image acquired based on the RGB color sp
model to the CIELAB color space model, a two-step conversion process was performed.
model
First, thetoRGB
thecolor
CIELABspacecolor
modelspace model, atotwo-step
was converted the CIEXYZ conversion
color spaceprocess was perform
model using
the color matching function of CIE [27], and the CIEXYZ color space model was convertedmodel u
First, the RGB color space model was converted to the CIEXYZ color space
thethe
to color matching
CIELAB function
color space model ofthrough
CIE [27],
theand the CIEXYZ
conversion formulacolor space
of CIE [28].model was conve
The color
space model conversion process is described in more detail in a study by Rossel
to the CIELAB color space model through the conversion formula of CIE [28]. The c et al. [26].
space model conversion process is described in more detail in a study by Rossel et al.
2.2. Test Apparatus
In this study, digital images were captured after creating a darkroom environment by
2.2. Test Apparatus
completely blocking external light sources other than the lighting used in the experiment.
In this
As shown study,
in Figure digital
3, the images
digital imaging were captured
equipment afterofcreating
consisted a darkroom
a digital camera, environm
lighting,
and
by completely blocking external light sources other than the lighting used ina the exp
a mold for soil samples. The digital camera, lighting, and mold were each fixed to
frame
ment.with an adjustable
As shown position.
in Figure Thedigital
3, the distance betweenequipment
imaging the imaging consisted
surface (soilofsample
a digital cam
surface) and the camera’s image sensor was kept constant at 500 mm with the focal length
lighting, and a mold for soil samples. The digital camera, lighting, and mold were e
fixed (the camera’s autofocus function was not used), and the lighting was 700 mm from
fixed
the to a frame
imaging surface.with an adjustable position. The distance between the imaging sur
(soil sample surface) and the camera’s image sensor was kept constant at 500 mm with
OR PEER REVIEW 5 of 17

Sensors 2023, 23, 296 5 of 16

focal length fixed (the camera’s autofocus function was not used), and the lighting was
700 mm from the imaging surface.

200mm

500mm

125mm

20mm

(a) (b)
Figure 3. (a) Schematic diagram
Figure 3. (a) and (b) photograph
Schematic diagram andof
(b)the digital image
photograph studio.image studio.
of the digital

The digital camera Theuseddigital

was camera
a Nikon used waswith
D850 a Nikon D850 with
an AF-S 50 mman AF-S
prime 50 lens
mm prime lens attached.
attached.
This device can acquire 45.75-million-pixel digital images using a full-frame CMOS sensor,
This device can acquire 45.75-million-pixel digital images using a full-frame CMOS sensor,
and minimizes camera shake due to influences, such as shutter operation. Two GODOX
and minimizes cameraLED shake
lights, due to influences,
SL100Bi, were used such as shutter
for lighting; operation.
based Two GODOX
on the imaging surface, the color
LED lights, SL100Bi,temperature
were usedcould for lighting;
be adjustedbased on the
from 2800 K toimaging
6000 K and surface, the color
the illuminance tem-
could be adjusted
perature could be adjusted from
up to 65,000 lux.2800 K tocontaining
The mold 6000 K and the the illuminance
soil sample was madecould be adjusted
of matte black plastic with
an inner
up to 65,000 lux. The molddiameter
containingand height of 125
the soil mm and
sample was 20 made
mm, respectively. A blackout
of matte black curtain was
plastic
with an inner diameter and height of 125 mm and 20 mm, respectively. A blackout curtain of light
placed on the bottom surface where the mold was placed to minimize the effect
reflected from objects other than the sample. In addition, image distortion was minimized
was placed on the bottom surface where the mold was placed to minimize the effect of
by arranging the image-taking surface in the vertical direction with the digital camera.
light reflected from objects other than the sample. In addition, image distortion was min-
imized by arranging 2.3.the
Soilimage-taking
Specimen surface in the vertical direction with the digital
camera. In this study, four soil samples (Jumunjin sand, Anseong soil, Yongin soil, and
Gwanak soil) with different colors were used, according to the location of collection and
2.3. Soil Specimen differences in the constituent minerals. Jumunjin sand is a silica-based sand collected
from Jumunjin Beach in Gangneung, Republic of Korea, and has a bright yellow color.
In this study, four soil samples
Anseong (Jumunjin
soil and Yongin sand,
soil are Anseong
weathered soil,
granite soilYongin
collectedsoil,
from and Gwa-
borrow pit located
nak soil) with different colorsand
in Anseong were used,respectively,
Yongin, accordingand to the location
Gwanak soil isofweathered
collection and dif-
granite soil collected
from Mt. minerals.
ferences in the constituent Gwanak inJumunjin
Seoul, Republic
sandofisKorea.
a silica-based sand collected from
In accordance with ASTM [29],
Jumunjin Beach in Gangneung, Republic of Korea, and has a bright mechanical sieve analysis
yellow was performed
color. Anseong to analyze
the particle size of four soil samples (see Figure 4). Jumunjin sand was classified as
soil and Yongin soil are weathered granite soil collected from borrow pit located in An-
poorly graded sand (SP) by the Unified Soil Classification System (USCS) and had an
seong and Yongin, respectively,
average effective andparticle
Gwanak soil is(D
diameter weathered granite soil collected from
50 ) of 0.58 mm. Anseong soil, Yongin soil, and
Mt. Gwanak in Seoul, Republic
Gwanak of Korea.
soil were classified as SM (silty sand) according to the USCS. The average effective
In accordance particle
with ASTM [29],(D
diameters mechanical sieve
50 ) of the three soil analysis was performed
samples (Anseong to analyze
soil, Yongin soil, and Gwanak
soil) were 1.05 mm, 0.96 mm, and 0.32 mm, respectively,
the particle size of four soil samples (see Figure 4). Jumunjin sand was classified as poorlyand the passing percentages of
200 sieves (the ratios of particles with a diameter of
graded sand (SP) by the Unified Soil Classification System (USCS) and had an average 0.074 mm or less) were 16.8%, 13.2%,
and 31.9%, respectively.
effective particle diameter (D50) of 0.58 mm. Anseong soil, Yongin soil, and Gwanak soil
were classified as SM (silty sand) according to the USCS. The average effective particle
diameters (D50) of the three soil samples (Anseong soil, Yongin soil, and Gwanak soil)
were 1.05 mm, 0.96 mm, and 0.32 mm, respectively, and the passing percentages of 200
sieves (the ratios of particles with a diameter of 0.074 mm or less) were 16.8%, 13.2%, and
Sensors 2022, 22, x FOR PEER REVIEW 6
Sensors 2023, 23, 296 6 of 16

100
90

Percent finer (%)

80
70
60
50
40 Jumunjin sand
30 Anseong soil
20 Yongin soil
10 Gwanak soil
0
10 1 0.1 0.01
Particle size (mm)
Figure4.4.Particle
Figure Particlesizesize distribution
distribution of theof the
test test specimens.
specimens.

Four
Foursoil samples
soil drieddried
samples at 110at 5 ◦ C ±for
± 110 5 more than
°C for 24 h than
more were compacted
24 h wereinto cylindri- into c
compacted
cal molds to prepare the samples for photography. The samples for photography each had
drical molds to prepare the samples for photography. The samples for photography
a dry density corresponding to 70% of the relative density of each sample (the dry densities
had
of a dry sand,
Jumunjin density corresponding
Anseong soil, Yongin to 70%
soil, and of the relative
Gwanak density
soil are 1.52 g/cmof eachg/cm
3 , 1.73 sample
3, (the
densities
1.71 g/cm3 ,of Jumunjin
and 1.73 g/cm sand, Anseong A
3 , respectively). soil, Yonginrod
compactor soil, and
with theGwanak soil are
same diameter as 1.52 g/
1.73inner
the g/cm 3 , 1.71 g/cm
diameter of the, mold
3 and 1.73 g/cmso, respectively).
was used 3 that each sample Awas
compactor rod with
homogeneously andthe sam
flatly
ameter compacted.
as the inner diameter of the mold was used so that each sample was homog
ously and flatly compacted.
2.4. Test Conditions
In this study, digital images were captured after simulating natural light conditions
2.4. Test Conditions
using lighting in a dark room completely blocked from external light sources. In this case,
In light
“natural this study, digital
conditions” images
refers were captured
to illuminance and colorafter simulating
temperature, natural
which light condit
are indica-
tors thatlighting
using quantitatively represent
in a dark roomthe lightness and
completely chromaticity,
blocked respectively,
from external lightofsources.
the light In this
source.
“natural Table 1 shows
light the illuminance
conditions” refers to and color temperature
illuminance of natural
and color light measured
temperature, which in are ind
Goyang-si, Gyeonggi-do, Republic of Korea from 15 February 2022 to 28 March 2022 using
tors that quantitatively represent the lightness and chromaticity, respectively, of the
Konica Minolta’s portable optical property measuring device CL-200A. The illuminance
source.
of naturalTable 1 shows
light was in the the illuminance
range of 15,540–65,040and lux,
color temperature
and of natural
showed a tendency light measure
to decrease
Goyang-si,
from morningGyeonggi-do,
to afternoon and Republic of Korea
during cloudy fromIn15addition,
weather. February the 2022
color to 28 March 2022 u
temperature
Konica
of naturalMinolta’s
light was inportable
the rangeoptical property
of 3590–5808 K andmeasuring
showed smaller device
valuesCL-200A. The illumin
close to sunset.
The measured color temperature showed a range similar to that
of natural light was in the range of 15,540–65,040 lux, and showed a tendencymeasured by Jeon et al. [30] to decr
for four seasons from April 2017 to April 2018.
from morning to afternoon and during cloudy weather. In addition, the color tempera
Based on the measurement results of the illuminance and color temperature of natural
of natural
light, light
the light was in the
conditions to berange
appliedof 3590–5808
when takingKdigital
and showed
images of smaller valueswere
soil samples close to su
The measured
determined colorintemperature
as shown Table 2. For the showed
four soilasamples
range similar
(Jumunjin to sand,
that measured
Anseong soil, by Jeon e
[30] forsoil,
Yongin fourand seasons
Gwanak from April
soil), 2017
digital to April
images were2018.
captured while changing the light
conditions 15 times (60 shots in total). Light with a target illuminance (15,000 lux, 35,000 lux,
50,000
Table lux, and 65,000 lux)
1. Illuminance and and
colorcolor temperature
temperature (3000 K,light
of natural 3800in K,Goyang-si,
4500 K, andGyeonggi-do,
5500 K) was Repub
irradiated onto the sample
Korea (15 February 2022–28using
March the2022).
LED lights. At this time, slight errors occurred in
the process of manually adjusting the illuminance and color temperature of the lighting
(average errors of illuminance and color temperature Direct were 1.57% and 1.05%, respectively).Color Temperatu
Datebefore taking
Therefore, Timean image Weatherof the soil sample, we Illuminance
measured the (lux)
illuminance and color
Light * (K)
temperature of the soil sample actually received from the image-taking surface through
02/15/2022 13:30 Clear O 65,040
using the optical property measuring device, and the measured values were used to analyze
5350
the02/16/2022
results. 17:00 Clear O 33,730 3762
02/17/2022 14:30 Clear O 47,580 5237
02/18/2022 10:30 Clear O 58,530 5353
02/20/2022 10:30 Clear O 60,200 5551
02/21/2022 14:30 Cloudy X 15,540 5808
02/22/2022 10:30 Clear O 60,100 5255
Sensors 2023, 23, 296 7 of 16

Table 1. Illuminance and color temperature of natural light in Goyang-si, Gyeonggi-do, Republic of
Korea (15 February 2022–28 March 2022).

Direct Color
Date Time Weather Illuminance (lux)
Light * Temperature (K)
02/15/2022 13:30 Clear O 65,040 5350
02/16/2022 17:00 Clear O 33,730 3762
02/17/2022 14:30 Clear O 47,580 5237
02/18/2022 10:30 Clear O 58,530 5353
02/20/2022 10:30 Clear O 60,200 5551
02/21/2022 14:30 Cloudy X 15,540 5808
02/22/2022 10:30 Clear O 60,100 5255
03/03/2022 10:30 Clear O 59,520 5312
03/08/2022 14:30 Cloudy X 16,700 5252
03/08/2022 14:30 Cloudy O 43,720 5122
03/11/2022 17:00 Cloudy O 28,720 3640
03/15/2022 14:30 Cloudy O 45,570 5605
03/22/2022 10:30 Clear O 49,200 5342
03/22/2022 17:00 Clear O 32,750 3811
03/27/2022 14:30 Clear X 16,210 5355
03/28/2022 17:00 Clear O 35,420 3590
* “O” represents the direct sunlight condition, whereas “X” represents the shaded condition.

Table 2. Lighting conditions for digital soil image acquisition.

Illuminance (lux) Color Temperature (K)

3000
15,000 4500
5500
3000
3800
35,000
4500
5500
3000
3800
45,000
4500
5500
3000
3800
65,000
4500
5500

Because digital camera settings (aperture value, shutter speed, ISO, white balance,
flash, etc.) have a great influence on digital image measurement results, it is necessary to
set appropriate values according to the shooting conditions and subjects. In this study, the
aperture, shutter speed, and ISO values were set to f/5, 1/1000 s, and 200, respectively,
considering the size and shooting distance of the soil sample. The white balance was fixed
at 5500 K and the picture was taken with the flash off, so the influence of the camera settings
on the image measurement results was excluded.

3. Results and Discussion

3.1. Digital Soil Images
The soil color change according to light conditions was analyzed based on the digital
image capture results of soil samples. Because the trend of soil color change according to
light conditions was similar regardless of the soil samples, digital images of Yongin soil
and Gwanak soil are shown as representative examples.
Sensors
Sensors 2022,
2022, 22,
22, xx FOR
FOR PEER
PEER REVIEW
REVIEW 88 of
of 17
17

Sensors 2023, 23, 296 8 of 16

light conditions was similar regardless of the soil samples, digital images of Yongin soil
and Gwanak soil are shown as representative examples.
Figures 5 and 6 show digital images of Yongin soil and Gwanak soil, respectively,
Figures 5 and 6 show digital images of Yongin soil and Gwanak soil, respectively,
taken under three different illuminance (15,000 lux, 35,000 lux, and 65,000 lux) and color
taken under three different illuminance (15,000 lux, 35,000 lux, and 65,000 lux) and color
temperature (3000 K, 4500 K, and 5500 K) conditions. Even though the identical soil sam-
temperature (3000 K, 4500 K, and 5500 K) conditions. Even though the identical soil samples
ples were photographed using the same camera settings, the soil color displayed on the
were photographed using the same camera settings, the soil color displayed on the image
image varied depending
varied greatly greatly depending on conditions.
on the light the light conditions. This as
This is because, is described
because, as described
above, when
above, when the incident light is changed according to the color expression principle, the
the incident light is changed according to the color expression principle, the reflected light
reflected light that determines the color
that determines the color also changed. also changed.

Figure 5. Digital images of Yongin soil.

Figure 5. Digital images of Yongin soil.

Figure 6.
Figure Digital images
6. Digital images of
of Gwanak
Gwanak soil.
soil.

The color temperature is a numerical representation of the chromaticity of the light

source using absolute temperature (K). The color temperature of a red light source is lower,
and the color temperature of a blue light source is higher. The illuminance of the light
Sensors 2023, 23, 296 9 of 16

source is a value indicating the intensity of light received by a specific surface; the higher
the illuminance, the brighter the light source. Each soil color also showed the same trend
regarding the changes in the light source according to color temperature and illuminance
(see Figures 5 and 6). That is, as the color temperature of the lighting increased, the soil’s
color became closer to blue than red, and as the illuminance increased, it became brighter.
The phenomenon of soil color change depending on lighting illustrates the limitations of
existing studies [16–23].
The above studies suggested a method of predicting soil property using soil color
obtained based on digital image processing, but were conducted based on images taken in
an indoor studio under constant light conditions. The suggested research results cannot be
applied in outdoor environments where the light conditions do not match those used in
the studies, or where it is practically impossible to control the light conditions. In order to
make the above research results more versatile and applicable in the field, a digital image
processing-based soil color analysis method that can consider irregular light conditions
(color temperature and illuminance) is required.
The next section describes the digital image processing performed to numerically
acquire the soil colors of the samples in two color space models (RGB and CIELAB). The
color change according to the light conditions was analyzed, and a color space model
capable of considering irregular light conditions was confirmed.

3.2. Soil Color Based on Digital Image Processing

In this study, digital image processing was performed in three major steps (region
of interest (ROI) setting, color space model conversion, and color extraction within the
region of interest). As shown in Figure 7, the ROI was set as a circle with a diameter of
100 mm that shared its center with the soil sample (the diameter of the soil sample was
125 mm). Colorimetric conversion was performed according to the procedure described in
Section 2. Referring to Gomez-Robledo et al. [22], the statistical mode (hereinafter referred
to as “mode”) was used as a representative value of soil color in the ROI. This is because
Garcia et al. [31] demonstrated that the mode is more suitable than the mean value for
the analysis of heterogeneous color images (here, soil images), thereby eliminating
Sensors 2022, 22, x FOR PEER REVIEW 10 of 17 the
undesirable effect of noise.

Figure7.
Figure 7. Region
Regionofofinterest (ROI)
interest (ROI)in digital image
in digital analysis:
image (a) Jumunjin
analysis: sand, (b)
(a) Jumunjin Anseong
sand, soil, (c) soil,
(b) Anseong
Yongin soil, and (d) Gwanak soil.
(c) Yongin soil, and (d) Gwanak soil.

Simple linear regression analysis was performed using light conditions (illuminance
and color temperature) as independent variables (explanatory variables) and soil color
intensity as dependent variables. Linear regression analysis is usually performed unless
there is clear evidence that a non-linear relationship (e.g., polynomial relationship, expo-
 Sensors 2023, 23, 296 10 of 16

Simple linear regression analysis was performed using light conditions (illuminance
and color temperature) as independent variables (explanatory variables) and soil color
intensity as dependent variables. Linear regression analysis is usually performed unless
there is clear evidence that a non-linear relationship (e.g., polynomial relationship, expo-
nential regression, logarithmic regression) exists between the independent and dependent
variables. There is no clear evidence to indicate a non-linear relationship between the light
condition and the soil color intensity; also, there is not enough data accumulated to perform
a non-linear regression analysis. As a performance indicator of the simple linear regression
model, the coefficient of determination (R2 ) that the ratio of the variance of the dependent
variable explained by the independent variable in the linear regression analysis using the
least squares method was used.
Figures 8 and 9 show the mode values of soil colors (R, G, and B) derived for the RGB
color display system with respect to the illuminance and color temperature, respectively,
of the light received by the soil. As shown in Figure 8, in all four soil samples, as the
illuminance increased, the RGB color intensity of the soil tended to increase. In the case of
green (G), which is known to have the highest correlation with brightness among the three
colors of RGB, the correlation coefficient with illuminance (R2 ) was higher than 0.91, but
the correlation coefficients with red (R) and blue (B) were relatively low. This is because
the correlation coefficient is relatively low compared to the illuminance, but the color
temperature also affects the soil’s colors (especially the intensity of red (R) and blue (B)). As
Sensors 2022, 22, x FOR PEER REVIEW can be seen in Figure 9, the red intensity of the soil sample photographed under lighting 11 of 1
with a low color temperature (red series) was high, and the blue intensity of the soil sample
photographed under the lighting with a high color temperature (blue series) was high.

Figure 8. Mode
Figure 8. Mode RGB RGB color
colorintensities
intensities according
according totothe
theilluminance
illuminance of light:
of light: (a) Jumunjin
(a) Jumunjin sand,sand, (b
(b) Anseong
Anseong soil,
soil, (c) (c) Yongin
Yongin soil,soil,
and and
(d)(d)Gwanak
Gwanak soil.
soil.

One interesting finding regarding the relationship between RGB color intensity and
illuminance is that the linear regression equations of the four soil samples with different
constituent minerals had similar slopes (see Figure 8). Because there were differences in
the soil color of the four samples, there were differences in the y-intercept of the linear
regression equation, but there was no significant difference in the degree of soil color change
(i.e., slope) according to the change in illuminance. Nevertheless, the soil color displayed
Sensors 2023, 23, 296 11 of 16

by the RGB color display system is affected by both illuminance and color temperature;
Figure
when8. using
Modethe RGB color
RGB intensities
color according
space model, to the
it would illuminance
be difficult of light:
to calibrate (or(a) Jumunjin
exclude) the sand, (b
Anseong soil, (c) Yongin soil, and (d) Gwanak
effects of illuminance and color temperature. soil.

Figure 9. Mode
Figure 9. ModeRGBRGBcolor
color intensities according
intensities according to the
to the colorcolor temperature
temperature of light:of(a)
light: (a) Jumunjin
Jumunjin sand, sand
(b) (b)
Anseong
Anseong soil,
soil,(c)
(c)Yongin soil,and
Yongin soil, and (d)(d) Gwanak
Gwanak soil.soil.

Figures 10 and 11 show the mode values of soil colors (L*, a*, and b*) derived for the
One interesting finding regarding the relationship between RGB color intensity an
CIELAB color display system for the illuminance and color temperature, respectively, of
illuminance is that by
the light received thethe
linear regression
soil. In equations
all four soil samples, L* ofshowed
the four soil samples
a high correlationwith
withdifferen
constituent
illuminance, minerals had
and a* and b* similar
showed aslopes (see Figure
high correlation with8). Because
color there(Rwere
temperature differences
2 > 0.90). In i
thecontrast,
soil color of the four
illuminance had samples,
little effectthere were
on a* and b*,differences in the y-intercept
and color temperature of the linea
had little effect
on L*. This
regression is becausebut
equation, the there
CIELAB wascolor
nospace model has
significant the characteristic
difference in the of expressing
degree of soil colo
change (i.e., slope) according to the change in illuminance. Nevertheless, thethesoil colo
colors by dividing them into “lightness” (L*-axis) and “chromaticity” (a*-b* plane). In
CIELAB color space model, L* is an index indicating the lightness of a color, so L*, the
displayed by the RGB color display system is affected by both illuminance and color tem
lightness of the soil color, increased as the illuminance increased. As the color temperature
perature;
increased when using the
and changed fromRGBred tocolor
blue, space
the soilmodel, it would
color moved be difficult
away from to calibrate
red and closer to (o
exclude)
blue, andthetheeffects of illuminance
color indicators a* (red and color
to green) temperature.
and b* (yellow to blue) gradually decreased
(see Figure 2). Furthermore, similar to the relationship between RGB color intensity and
illuminance, the linear regression slopes of (1) L* and illuminance, and (2) a*, b* and color
temperature were similar regardless of the type of soil sample. Because the soil colors
of the four samples were different, there were differences in the y-intercept of the linear
regression equation, but there was no significant difference in the degree of change in soil
color according to changes in illuminance and color temperature.
In summary, among the color components, L* was affected only by illuminance,
but a* and b* were affected only by color temperature, and the correlation was high.
Therefore, based on the relationships between (1) illuminance and color temperature and
(2) illuminance and color temperature obtained based on the CIELAB color space model,
the effects of illuminance and color temperature on soil color can be calibrated (or excluded).
It is thought that this goal can be achieved using our novel calibration method, which is
described in more detail in the following subsection.
 creased (see Figure 2). Furthermore, similar to the relationship between RGB color inten
sity and illuminance, the linear regression slopes of (1) L* and illuminance, and (2) a*, b
and color temperature were similar regardless of the type of soil sample. Because the soi
colors of the four samples were different, there were differences in the y-intercept of th
Sensors 2023, 23, 296 linear regression equation, but there was no significant difference in the degree of chang
12 of 16
in soil color according to changes in illuminance and color temperature.

Sensors 2022, 22, x FOR PEER REVIEW 13 of 1

Figure 10.10.
Figure Mode
ModeCIELAB
CIELAB color intensities
color intensities according
according to illuminance
to the the illuminance of(a)
of light: light: (a) Jumunjin
Jumunjin sand, sand
(b) Anseong
(b) Anseongsoil,
soil,(c)
(c)Yongin soil,and
Yongin soil, and(d)(d) Gwanak
Gwanak soil.soil.

Figure 11.11.
Figure Mode
ModeCIELAB
CIELABcolor intensitiesaccording
color intensities according to color
to the the color temperature
temperature of light:of
(a)light: (a) Jumunjin
Jumunjin
sand, (b) (b)
sand, Anseong
Anseongsoil,
soil,(c)
(c) Yongin soil,and
Yongin soil, and
(d)(d) Gwanak
Gwanak soil. soil.

3.3. Digital Soil Image Calibration Method

In summary, among the color components, L* was affected only by illuminance, bu
Through digital image analysis of soil samples captured under various light conditions,
a* and b* were affected only by color temperature, and the correlation was high. Therefore
there is a linear relationship between the soil color displayed based on the CIELAB color
based on the relationships between (1) illuminance and color temperature and (2) illumi
nance and color temperature obtained based on the CIELAB color space model, the effect
of illuminance and color temperature on soil color can be calibrated (or excluded). It i
thought that this goal can be achieved using our novel calibration method, which is de
scribed in more detail in the following subsection.
Sensors 2023, 23, 296 13 of 16

space model and the illuminance and color temperature of the light received by the soil, as
follows (see Figures 10 and 11):
L∗ = a L I + f L , (1)
a∗ = a a T + f a , (2)
∗
b = ab T + f b . (3)
Here, I and T denote the illuminance and color temperature, respectively, of the light
received by the soil. aL and fL are the slope and intercept of the linear regression equation
between the L* value of the soil color and illuminance, respectively; aa and fa are the slope
and intercept of the linear regression equation between the a* value of the soil color and
color temperature, respectively; and ab and fb are the slope and intercept of the linear
regression between the b* value of the soil color and the color temperature. The digital soil
image calibration procedure through the above formula is as follows:
1. The target soil is sequentially irradiated by various configurations of light with four
or more different illuminances and color temperatures, and digital images are taken.
At this time, the light intensity and color temperature (i.e., I and T) of the target soil
are recorded;
2. From the digital soil images, the soil color (i.e., L*, a*, and b*) of the target soil is
obtained based on the CIELAB color space model;
3. A linear regression analysis of the relationship between the light conditions and soil
color is plotted on a two-dimensional plane (I-L* plane, T-a* plane, and T-b* plane);
using Equations (1)–(3), the slope (aL , aa , and ab ) and intercept (fL , fa , and fb ) are found.
Through the relational expression between the light condition and soil color deter-
mined in this way, the soil color of the target soil can be obtained under a given light
condition (substituting the given light condition into Equations (1)–(3)).
Table 3 shows the slope of the linear regression equation between L* and illuminance,
as well as that between a* and b* and color temperature, which showed a particularly
significant correlation coefficient among the relationships between light conditions and
soil color (see Figures 10 and 11). As mentioned earlier, the slopes of the linear regression
equations in Figures 10 and 11 were similar regardless of the type of soil sample. When the
average value of the linear regression slope is substituted into the slopes (i.e., aL , aa , and
ab ) of Equations (1)–(3), the linear relationship between soil color and light conditions is
as follows:
L∗ = 0.0008I + f L , (4)
a∗ = −0.0126T + f a , (5)
b∗ = −0.0088T + f b . (6)

Table 3. Slope of linear regression lines between lighting conditions (illuminance and color tempera-
ture) and CIELAB-based soil color.

Soil Type aL aa ab
Jumunjin sand 0.0008 −0.0125 −0.0088
Anseong soil 0.0008 −0.0125 −0.0086
Yongin soil 0.0008 −0.0127 −0.0086
Gwanak soil 0.0008 −0.0128 −0.0091

Equations (4)–(6) are relational expressions between light conditions and soil color
that can be applied to soils similar to the samples used in this study (silica-based sand or
weathered granite soil). Through these equations, the relationship between light conditions
and soil color can be determined from just one digital image capture. The procedure is
as follows:
Sensors 2023, 23, 296 14 of 16

1. Take a digital image of the target soil, and record the light intensity and color temper-
ature (i.e., I and T) of the target soil;
2. Obtain the soil color (i.e., L*, a*, and b*) of the target soil based on the CIELAB color
space model;
3. Substitute in Equations (4)–(6) to find the intercepts (fL , fa , and fb ).
By substituting the given illuminance and color temperature into the relationship
between light conditions and soil color determined through this process, the soil color
under specific light conditions can be calculated.
As mentioned earlier, previous studies [16–23] proposed a correlation between soil
color and soil property based on digital images taken in an indoor studio under constant
light conditions. It should be noted that if the light conditions are different, the soil color
will be different, even for the same soil. Thus, to utilize the previously proposed correlation
between soil color and soil property, it is necessary to obtain the soil color by creating the
same light conditions as in previous studies and then capturing digital images. However, it
is difficult to create light conditions identical to those in the previous studies in an indoor
studio, and in particular, it is virtually impossible to control the light conditions in an
outdoor case.
From this point of view, the proposed soil color calibration method has great signifi-
cance. If Equations (1)–(3) or Equations (4)–(6) are determined based on the digital image
of a soil sample taken under arbitrary light conditions, the soil color under a specific light
condition can be calculated. First, to establish the correlation between soil color and soil
property, the soil colors (i.e., L*, a*, and b*) obtained based on the CIELAB color space
model are used. At this time, the illuminance and color temperature of the light received
by the soil sample are recorded. Then, to predict the soil properties, Equations (1)–(3) or
Equations (4)–(6) are determined using the digital images of soil samples taken under arbi-
trary light conditions. The soil color is obtained by substituting the specific light conditions
(illuminance and color temperature of light received by soil samples when establishing the
correlation between soil color and soil property) into the aforementioned formulas; this is
used to predict the soil property.
According to the above process, it is possible to calibrate the soil color photographed
under irregular light conditions to the corresponding color under the desired light condition.
It is expected that this approach will allow soil color to be used as a reliable indicator for
predicting soil properties, overcoming the limitations of existing studies resulting from
variation in soil color due to light conditions.

4. Conclusions
In this study, a new method was proposed to calibrate the color of soil, which changes
according to different light conditions. Digital images were captured while changing the
light conditions 15 times for 4 soil samples. The soil color of the samples photographed
through digital image processing was extracted according to two color space models
(RGB and CIELAB), and the color change according to the light conditions was analyzed.
Through this process, the following conclusions were reached.
1. Even when the uniformly formed soil samples were photographed with the same
camera settings, the soil color varied greatly depending on the color temperature and
illuminance of the light received by the soil. Most existing digital image processing-
based soil color analysis studies were performed under constant light conditions,
making them less applicable to outdoor environments with different light conditions,
or environments in which control of light conditions is virtually impossible;
2. Color analysis based on the RGB color space model indicated that the color parameters
(R, G, and B) increased as the illuminance increased. Among these parameters,
green (G), which is known to have the highest correlation with brightness, had a very
high correlation with illuminance, but red (R) and blue (B) showed relatively low
correlations due to the effect of color temperature. It was concluded that it would
be difficult to exclude (or calibrate) the influence of light conditions through this
Sensors 2023, 23, 296 15 of 16

approach, as the soil color displayed by the RGB color display system is partially
affected by both illuminance and color temperature;
3. The soil color analysis based on the CIELAB color space model showed that L* was
affected only by illuminance, but a* and b* were affected only by color temperature,
and the correlation was high. This is because L* represents the lightness of a color and
a* and b* are indices representing the chromaticity. Using the relationship between
illuminance and color temperature, as well as that between illuminance and color
temperature, obtained based on the CIELAB color indicator, it is possible to calibrate
(or exclude) the effect of illuminance and color temperature on soil color;
4. Using the linear regression equation of soil color and light conditions based on the
CIELAB color space model, a method for calibrating soil color was proposed. Through
the proposed method, the soil color of soil samples photographed under arbitrary
light conditions can be corrected to the soil color under the desired light conditions.
By excluding the effect of light conditions on soil color using the method proposed in
this study, soil color can be used as a reliable indicator for predicting soil properties, such
as constituent minerals, organic matter content, water content, and ion concentration. The
proposed method makes the soil color obtained from the digital images more versatile and
applicable in the field; thus, digital image processing will be a simple, fast, and inexpensive
method of on-site analysis for land and crops. However, the above results are derived
based on digital images taken for four soil samples (silica-based beach sand and weathered
granite soil). To make the conclusion more convincing, the proposed soil color calibration
process needs to be supplemented by acquiring soil color from different soil samples (with
different or similar mineral composition). It is necessary to analyze the correlation between
soil color and light conditions according to the framework established in this study using
more diverse soil samples.

Author Contributions: Conceptualization, S.-H.B.; methodology, S.-H.B.; software, S.-H.B. and

J.-S.J.; validation, T.-Y.K.; investigation, S.-H.B. and T.-Y.K.; experiment, S.-H.B., K.-H.P., J.-S.J. and
T.-Y.K.; writing—original draft preparation, S.-H.B.; writing—review and editing, S.-H.B. and T.-Y.K.;
visualization, S.-H.B.; supervision, K.-H.P.; project administration, K.-H.P. and T.-Y.K.; funding
acquisition, T.-Y.K. All authors have read and agreed to the published version of the manuscript.
Funding: Research for this paper was carried out under the KICT Research Program
(project no. 20220173-001, Geotechnical research for disaster management and new growth in the
future geotechnical industry) funded by the Ministry of Science and ICT.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Acknowledgments: Construction Engineering Research Institute at Hankyong National University
provided research facilities for this work.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Webster, R.; Butler, B.E. Soil classification and survey studies at Ginninderra. Soil Res. 1976, 14, 1–24. [CrossRef]
2. Ibanez-Asensio, S.; Marques-Mateu, A.; Moreno-Ramon, H.; Balasch, S. Statistical relationships between soil colour and soil
attributes in semiarid areas. Biosyst. Eng. 2013, 116, 120–129. [CrossRef]
3. Han, P.; Dong, D.; Zhao, X.; Jiao, L.; Lang, Y. A smartphone-based soil color sensor: For soil type classification. Comput. Electron.
Agric. 2016, 123, 232–241. [CrossRef]
4. Munsell Soil Colour Charts; The Munsell Soil Colour Book; Munsell Colour Company Inc.: Newburgh, NY, USA, 1992.
5. ASTM. Standard Practice for Specifying Color by the Munsell System; ASTM International D: West Conshohocken, PA, USA, 2014;
pp. 1514–1535.
6. Torrent, J.; Barron, V. Laboratory measurement of soil color: Theory and practice. Soil Color 1993, 31, 21–33.
7. Baillie, I.C. Soil survey staff 1999, soil taxonomy. Soil Use Manag. 2001, 17, 57–60. [CrossRef]
Sensors 2023, 23, 296 16 of 16

8. Schad, P.; Spaargaren, O. World Reference Base for Soil Resources 2006: A Framework for International Classification, Correlation and
Communication; Food and Agriculture Organization of the United Nations (FAO): Rome, Italy, 2006.
9. Thompson, J.; Pollio, A.; Turk, P. Comparison of Munsell Soil Color Charts and the GLOBE soil color book. Soil Sci. Soc. Am. J.
2013, 77, 2089–2093. [CrossRef]
10. Mouazen, A.M.; Karoui, R.; Deckers, J.; De Baerdemaeker, J.; Ramon, H. Potential of visible and near-infrared spectroscopy to
derive colour groups utilising the Munsell soil colour charts. Biosyst. Eng. 2007, 97, 131–143. [CrossRef]
11. Berns, R. Billmeyer and Saltzman’s Principles of Color Technology; John Wiley & Sons: Hoboken, NJ, USA, 2019.
12. Kirillova, N.P.; Vodyanitskii, Y.N.; Sileva, T.M. Conversion of soil color parameters from the Munsell System to the CIE-L*a*b*
System. Eurasian Soil Sci. 2015, 48, 468–475. [CrossRef]
13. Aydemir, S.; Keskin, S.; Drees, L.R. Quantification of soil features using digital image processing (DIP) techniques. Geoderma 2004,
119, 1–8. [CrossRef]
14. Rossel, R.A.; Fouad, Y.; Walter, C. Using a digital camera to measure soil organic carbon and iron contents. Biosyst. Eng. 2008,
100, 149–159. [CrossRef]
15. O’Donnell, T.K.; Goyne, K.W.; Miles, R.J.; Baffaut, C.; Anderson, S.H.; Sudduth, K.A. Determination of representative elementary
areas for soil redoximorphic features identified by digital image processing. Geoderma 2011, 161, 138–146. [CrossRef]
16. Persson, M. Estimating surface soil moisture from soil color using image analysis. Vadose Zone J. 2005, 4, 1119–1122. [CrossRef]
17. Zanetti, S.S.; Cecilio, R.A.; Alves, E.G.; Silva, V.H.; Sousa, E.F. Estimation of the moisture content of tropical soils using colour
images and artificial neural networks. Catena 2015, 135, 100–106. [CrossRef]
18. Santos, J.F.; Silva, H.R.; Pinto, F.A.; Assis, I.R.D. Use of digital images to estimate soil moisture. Rev. Bras. Eng. Agric. E Ambient.
2016, 20, 1051–1056. [CrossRef]
19. Park, J.S. Soil Classification and Characterization Using Unmanned Aerial Vehicle and Digital Image Processing. Ph.D. Thesis,
Seoul National University, Seoul, Republic of Korea, 2017.
20. Kim, D.K. Prediction on Physical Properties of Soil Based on Deep Learning using Digital Image Processing. Ph.D. Thesis,
Seoul National University, Seoul, Republic of Korea, 2020.
21. Zhu, Y.; Wang, Y.; Shao, M. Using soil surface gray level to determine surface soil water content. Sci. China Earth Sci. 2010,
53, 1527–1532. [CrossRef]
22. Gomez-Robledo, L.; Lopez-Ruiz, N.; Melgosa, M.; Palma, A.J. Using the mobile phone as munsell soil-colour sensor: An
experiment under controlled illumination conditions. Comput. Electron. Agric. 2013, 99, 200–208. [CrossRef]
23. Moonrungsee, N.; Pencharee, S.; Jakmunee, J. Colorimetric analyzer based on mobile phone camera for determination of available
phosphorus in soil. Talanta 2015, 136, 204–209. [CrossRef] [PubMed]
24. Billmeyer, F.W.; Saltzman, M. Principles of Color Technology; Wiley-Interscience: New York, NY, USA, 1981.
25. Wyszecki, G.; Stiles, W.S. Color Science: Concepts and Methods, Quantitative Data and Formulae; Wiley: New York, NY, USA, 1982.
26. Rossel, R.A.; Minasny, B.; Roudier, P.; McBratney, A.B. Colour space models for soil science. Geoderma 2006, 133, 320–337.
[CrossRef]
27. Commission Internationale de l’Eclairage (CIE). CIE Proceedings; Cambridge University Press: Cambridge, UK, 1931.
28. Commission Internationale de l’Eclairage (CIE). Recommendations on Uniform Color Spaces, Color Difference Equations, and Psychome-
tric Color Terms; CIE Publication: Vienna, Austria, 1978.
29. ASTM. Standard Test Methods for Particle-Size Distribution (Gradation) of Soils Using Sieve Analaysis; ASTM International: West
Conshohocken, PA, USA, 2009; Volume D6913-04.
30. Jeon, G.W.; Oh, S.T.; Lim, J.H. Algorithm for Judging Anomalies Using Sliding Window to Reproduce the Color Temperature
Cycle of Natural Light. J. Korea Multimed. Soc. 2021, 24, 30–39.
31. Garcia, P.A.; Huertas, R.; Melgosa, M.; Cui, G. Measurement of the relationship between perceived and computed color differences.
JOSA A 2007, 24, 1823–1829. [CrossRef] [PubMed]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.