2
2018). An alternative approach is to obtain the representa-
tive strength of coals by determining the megascopic coal
lithotypes in the field based on the coal brightness profile
by experienced geologists or mining engineers.
Stopes (1919) was the first researcher to introduce the
concept of the megascopic coal lithotype identification
with the lithotype nomenclature. Traditionally megascopic
character in banded bituminous coal is described composi-
tionally in terms of the abundance and distribution of the
four macro-lithotypes: vitrain, clarain, durain, and fusain.
Over the years, other coal description systems based on the
varying degrees of visible coal brightness and banding were
developed from Stope’s work. A simpler terminology and
approach have been adopted in Australia, which uses the
terms bright, banded, dull and fibrous, with designations
for banded bright and banded dull. The use of the coal
brightness methodology has gained widespread acceptance
in Australia. The brightness profile logging of coal core is an
accepted standard and is universally performed at Australian
mines. However, such a system of logging coal core is not
widely used in the United States. Rusnak (2017) conducted
a total of 1,000 uniaxial compressive strength tests and
440 indirect tensile strength tests on cores that were col-
lected from southern West Virginia and were logged with
the megascopic coal lithotype nomenclature. The statistical
analysis shows that the mechanical properties are correlated
to the lithotype. The correlation between coal strength and
megascopic coal lithotype was further confirmed by Rashed
et al. (2018) when using Schmidt hammer to estimate the
intact coal strength. The representative strength for bright
coal, banded bright coal, banded dull coal, and dull coal
was estimated in the study, making it convenient to deter-
mine the intact coal strength.
However, the megascopic coal lithotype is normally
determined by geologists or mining engineers in the field
or based on drilled cores. A number of problems arise from
these manual methods and the most important one is the
error from operator subjectivity. In order to overcome this
problem, Yu et al. (1997) employed a window filtering
method to analyze coal textural information acquired from
banded bituminous coal seams in the field. They tried to
discriminate the coal lithotypes accurately and automati-
cally by conducting statistical analysis on the sliding win-
dows captured from the coal images. However, there are
significant developments in photography technology and
image processing technique in the last two decades. The
purpose of this study is to use the advanced image process-
ing techniques to characterize megascopic coal lithotypes
based on the study of Yu et al. (1997).
As pointed out by Rashed et al. (2018), the coal bright-
ness profile infers a measure of volumetric cleat density,
where bright coal (BC) is characterized with fine cleat spac-
ing on the millimeter scale, and dull coal (DC) has wider
cleat spacing on the centimeter scale. Due to the presence
of dense cleats, it can be expected that the pixel values
change frequently and regularly, and that the difference in
cleat density is potentially reflected in the variation of pixel
values. Thus, it is reasonable to select an image processing
technique describing the relationship between pixels for
the textural analysis and lithotype classification. The gray
level co-occurrence matrix (GLCM) is such a method that
is widely used to describe image texture/patterns and to
extract features that are useful in various image processing
tasks such as image segmentation, classification, and fea-
ture extraction. Texture refers to the visual properties that
describe how rough, smooth, or patterned an area appears
within an image. GLCM is used to characterize and quan-
tify the texture or patterns in an image. The GLCM method
has been used to classify coal and rock where anthracite coal
and shale blocks were collected from an underground coal
mine and photos were taken in the laboratory and analyzed
with the GLCM method (Sun and Su 2013). Singh et al.
(2019) analyzed six different rock micro-CT images with
GLCM and revealed that every rock has its own GLCM
signature depending on the typical variations of the gray-
level intensities.
GLCM INTRODUCTION AND
APPLICATION
GLCM Introduction
GLCM is a statistical method of examining the image tex-
ture considering the spatial relationship of pixels. It is also
known as the gray level spatial dependence matrix, a matrix
that is defined over an image to be the distribution of the
co-occurring pixel values at a given offset. In simple terms,
a GLCM is a 2-dimensional matrix that contains informa-
tion about how often a pair of pixels with a specific inten-
sity (or “gray level”) appear in a certain orientation relative
to each other in an image. The orientation is specified by a
direction angle, such as 0, 45, 90, or 135 degrees, and the
specific gray level values for the pixel pair are used as indices
in the matrix. The value stored in the matrix at a given gray
level is the number of times the pair appears in the specified
orientation. Statistical measures can then be extracted from
the matrix, including contrast, correlation, dissimilarity,
energy, and homogeneity.
Contrast: measures the local variations or differences in
intensity between neighboring pixels. It is calculated as the
2018). An alternative approach is to obtain the representa-
tive strength of coals by determining the megascopic coal
lithotypes in the field based on the coal brightness profile
by experienced geologists or mining engineers.
Stopes (1919) was the first researcher to introduce the
concept of the megascopic coal lithotype identification
with the lithotype nomenclature. Traditionally megascopic
character in banded bituminous coal is described composi-
tionally in terms of the abundance and distribution of the
four macro-lithotypes: vitrain, clarain, durain, and fusain.
Over the years, other coal description systems based on the
varying degrees of visible coal brightness and banding were
developed from Stope’s work. A simpler terminology and
approach have been adopted in Australia, which uses the
terms bright, banded, dull and fibrous, with designations
for banded bright and banded dull. The use of the coal
brightness methodology has gained widespread acceptance
in Australia. The brightness profile logging of coal core is an
accepted standard and is universally performed at Australian
mines. However, such a system of logging coal core is not
widely used in the United States. Rusnak (2017) conducted
a total of 1,000 uniaxial compressive strength tests and
440 indirect tensile strength tests on cores that were col-
lected from southern West Virginia and were logged with
the megascopic coal lithotype nomenclature. The statistical
analysis shows that the mechanical properties are correlated
to the lithotype. The correlation between coal strength and
megascopic coal lithotype was further confirmed by Rashed
et al. (2018) when using Schmidt hammer to estimate the
intact coal strength. The representative strength for bright
coal, banded bright coal, banded dull coal, and dull coal
was estimated in the study, making it convenient to deter-
mine the intact coal strength.
However, the megascopic coal lithotype is normally
determined by geologists or mining engineers in the field
or based on drilled cores. A number of problems arise from
these manual methods and the most important one is the
error from operator subjectivity. In order to overcome this
problem, Yu et al. (1997) employed a window filtering
method to analyze coal textural information acquired from
banded bituminous coal seams in the field. They tried to
discriminate the coal lithotypes accurately and automati-
cally by conducting statistical analysis on the sliding win-
dows captured from the coal images. However, there are
significant developments in photography technology and
image processing technique in the last two decades. The
purpose of this study is to use the advanced image process-
ing techniques to characterize megascopic coal lithotypes
based on the study of Yu et al. (1997).
As pointed out by Rashed et al. (2018), the coal bright-
ness profile infers a measure of volumetric cleat density,
where bright coal (BC) is characterized with fine cleat spac-
ing on the millimeter scale, and dull coal (DC) has wider
cleat spacing on the centimeter scale. Due to the presence
of dense cleats, it can be expected that the pixel values
change frequently and regularly, and that the difference in
cleat density is potentially reflected in the variation of pixel
values. Thus, it is reasonable to select an image processing
technique describing the relationship between pixels for
the textural analysis and lithotype classification. The gray
level co-occurrence matrix (GLCM) is such a method that
is widely used to describe image texture/patterns and to
extract features that are useful in various image processing
tasks such as image segmentation, classification, and fea-
ture extraction. Texture refers to the visual properties that
describe how rough, smooth, or patterned an area appears
within an image. GLCM is used to characterize and quan-
tify the texture or patterns in an image. The GLCM method
has been used to classify coal and rock where anthracite coal
and shale blocks were collected from an underground coal
mine and photos were taken in the laboratory and analyzed
with the GLCM method (Sun and Su 2013). Singh et al.
(2019) analyzed six different rock micro-CT images with
GLCM and revealed that every rock has its own GLCM
signature depending on the typical variations of the gray-
level intensities.
GLCM INTRODUCTION AND
APPLICATION
GLCM Introduction
GLCM is a statistical method of examining the image tex-
ture considering the spatial relationship of pixels. It is also
known as the gray level spatial dependence matrix, a matrix
that is defined over an image to be the distribution of the
co-occurring pixel values at a given offset. In simple terms,
a GLCM is a 2-dimensional matrix that contains informa-
tion about how often a pair of pixels with a specific inten-
sity (or “gray level”) appear in a certain orientation relative
to each other in an image. The orientation is specified by a
direction angle, such as 0, 45, 90, or 135 degrees, and the
specific gray level values for the pixel pair are used as indices
in the matrix. The value stored in the matrix at a given gray
level is the number of times the pair appears in the specified
orientation. Statistical measures can then be extracted from
the matrix, including contrast, correlation, dissimilarity,
energy, and homogeneity.
Contrast: measures the local variations or differences in
intensity between neighboring pixels. It is calculated as the