Upload
vanquynh
View
223
Download
4
Embed Size (px)
Citation preview
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
968
Effect of particle size and shape on the grain-size distribution using
Image analysis Seracettin Arasan
1, Suat Akbulut
1, A.Samet Hasiloglu
2
1 – Department of Civil Engineering, Engineering Faculty,
Ataturk University, Erzurum, Turkey
2 – Department of Computer Engineering, Engineering Faculty,
Ataturk University, Erzurum, Turkey
doi:10.6088/ijcser.00202010083
ABSTRACT
Grain-size distribution that is one of the base properties of soils and soil classification
systems gives idea about engineering properties of soils. Determination of grain-size
distribution derived from mechanical method (sieving) is time consuming and difficult.
Hence, the image analysis methods for determination of grain-size distribution have been
also investigated by several researchers. In that research, the grain-size distributions for
ten soil samples with different sizes and shapes are determined with image analysis
methods. The effects of particle shape (i.e., elongated, flat, spherical) and volume
computation techniques (cube, cylinder, ellipsoid, and modified ellipsoid) on grain-size
distribution were also researched. The results of the research indicated that the shape of
particles significantly affected grain-size distribution. Additionally, modified ellipsoid
method is determined as the best volume computation method.
Keywords: Grain-size distribution, image analysis, soil, particle shape
1. Introduction
The grain-size distributions (GSD) are one of the basic and most important properties of
soil. It is primarily used for soil classification and provided a first-order estimate of other
soil engineering properties such as permeability, shear strength, and compressibility. In
practice, the GSD of the full size spectrum of soil grains is determined by integrating data
obtained from two inherently dissimilar tests, mechanical sieving for the coarse grained
soil fraction and hydrometer tests for the fine fraction. In sieve analysis, the particle size
is characterized by a single linear dimension representing the minimum square sieve
aperture that which the particle just passed through (Ghalib and Hryciw, 1999). The
results of sieving are dependent upon the shape of the particles (Mora et al., 1998; Kwan
et al., 1999; Mora and Kwan, 2000). Particles passing through a sieve can actually have
one dimension that is larger than the size of the sieve apertures. From Fig. 1(a), it can be
seen that an elongated particle that has its length greater than the aperture size can pass
through the sieve without any difficulties. Therefore, the sieve aperture size is a measure
of the lateral dimensions of the particles only. A relatively flat particle can pass through
the sieve aperture, which is square in shape, diagonally as shown in Fig. 1(b). As a result,
the breadth of a particle passing through a sieve can also be greater than the sieve size,
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
969
although it has to be smaller than the diagonal length of the sieve aperture (Mora et al.,
1998).
(a) (b)
Figure 1: (a) An elongated particle passing through a sieve aperture. (b) Plan view of a
flat particle passing through a sieve aperture. (Kwan et al., 1999)
The determination of grain-size distribution with mechanical method (sieving) is difficult
and it takes long time. Hence, image analysis has been used by several researchers to
determine the size distribution, particle shape and surface texture of aggregates, both in
two and three dimensions (Mora et al., 1998; Kwan et al., 1999; Mora and Kwan, 2000;
Maerz and Lusher, 2001; Rao et al., 2001; Taylor, 2002; Maerz, 2004; Fernlund,
2005a,b,c; Tutumluer et al., 2005; Lira and Pina, 2007). Although soil particles are 3-D in
nature, most of the image analysis techniques treat particles as two-dimensional objects
because only the two-dimensional projection of the particles is captured and measured.
However, for obtaining 3D images, many researchers used laser scanners (Lanaro and
Tolppanen, 2002; Tolppanen et al., 2002; Lee et al., 2007) and flatbed color scanner (Lira
and Pina, 2007), some are also used autofocus microscope (Chandan et al., 2004; Masad
et al., 2005) and optical interferometer (Alshibli and Alsaleh, 2004), and also the others
used two different perspective images (Kuo et al., 1996; Rao and Tutumluer, 2000;
Maerz, 2004; Fernlund, 2005a). These methods analyzed only one particle at a time,
however, and significantly increased the complexity compared to two-dimensional
analysis.
When plotting grain-size curves by sieve analysis, the particle size is presented with
respect to the percent cumulative mass of particles. However, image analysis could not
measure particle mass so results could be only presented in percent of particles (Fernlund,
1998; Andriani and Walsh, 2002; Fernlund, 2005a, b, c) or percent of area (Mora et al.,
1998; Mora and Kwan, 2000; Lira and Pina, 2007). Several researchers have also claimed
that more accurate volume and mass determinations are necessary for accurate plotting of
grain-size curves (Taylor, 2002; Maerz, 2004; Tutumluer et al., 2005; Fernlund et al.,
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
970
2007). A great number of studies dealing with volume computations of particles are
based on stereology, to convert data from areas (Barbery, 1991; King, 1984; Lin et al.,
1987; Vallebuona et al., 2003). These methods are applied for obtaining parameter
estimates (mean, variance and skewness) in three dimensions from two-dimensional data.
Bozdag and Karpuz (1997) also estimated the 3D size distribution of soil particles from
2D measures such as length and area by using the method proposed by King (1984) and
the moment method. The other studies are focused on geometrical computations (Mertens
and Elsen, 2006; Lira and Pina, 2007; Fernlund et al., 2007).
The aims of this paper are to investigate the effect of particle size and shape (i.e.,
elongated, flat, spherical) on the GSD, to compare the image analysis methods (top view,
front view, and volume) and to determine the more accurate method for volume
computations. For this purpose, ten soil samples with different shapes and sizes are used.
The results from the research were compared with those of the mechanical analysis.
2. Materials and Methods
2.1. Soil Samples
In this research, four different size fractions (i.e., 2mm-4,75mm, 4,75mm-8mm, 2mm-
8mm, and 8mm-16 mm) of a river soil were used to determine grain-size distribution.
Additionally, 4,75mm-8mm and 8mm-16mm size fractions were used for investigation of
particle shapes effect on GSD. For this purpose, particles were selected by hand as flat,
elongated, spherical, and mixed as mentioned below section. The classification of soils is
also mentioned in Table 1 according to the Unified Soil Classification System-USCS.
Table 1: The size fractions and classes of the soils used in this study
Size
Fraction
Soil
Classification
Particle
Shape
S1 2 mm-4,75 mm SP Mixed
S2 2 mm-8 mm SP Mixed
S3 4,75 mm-8 mm GP Mixed
S4 8 mm-16 mm GP Mixed
S5 4,75 mm-8 mm GP Elongated
S6 4,75 mm-8 mm GP Flat
S7 4,75 mm-8 mm GP Spherical
S8 8 mm-16 mm GP Elongated
S9 8 mm-16 mm GP Flat
S10 8 mm-16 mm GP Spherical
2.2. Particle Shapes
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
971
Quantification of form (particle shape) requires the measurement of the length (L),
breadth (I) and thickness (S) of a particle. A number of different identifications have been
used historically. Most previous authors have firstly identified the dimension L, defined
as the maximum caliper dimension, and have then measured I and S dimensions
orthogonal to L (Blott and Pye, 2008). Zingg (1935) indicated that the particles classified
four base classes as mentioned in Figure 2(a) with using above measurement system.
Also, Sneed and Folk (1958) considered Zingg’s diagram to be inadequate in that it
contained only four form classes which divided the field of form variation very unequally.
They suggested that only three end-members limit the system of dimensional variation: a
prolate spheroid with one long axis and two short ones (L > I = S), an oblate spheroid
with two long axes and one short one (L = I > S), and a sphere with all axes equal (L = I
= S). A number of different notations have been used historically to refer to these three
classes, including, rod, disk and cube (Krumbein, 1941), columnar, flat and spherical
(Blott and Pye, 2008). Sneed and Folk (1958)’s three classes are also used in this study
with different notation as elongated, flat, and spherical, respectively (Figure 2b).
(a) (b)
Figure 2: (a)Four different particle shapes (Zingg, 1935) (b) Three different particle
shapes (Sneed and Folk, 1958)
2.3. Imaging System
A high quality image of the particles is needed before any image processing can be
performed. In this study, the equipment for taking pictures of particles is set up by
mounting a camera on a photographic stand as shown in Figure 3(a), adjusting the height
of the camera to obtain a sufficiently large measurement area on the sample tray, and
adjusting the light sources so that there is no shading of any object placed on the sample
tray. A white cotton cloth is laid on the sample tray before the particles are put in for
obtaining better contrast between the particle and the background pixels. A Nikon D80
Camera and Micro 60 mm objective manufactured by Nikon were used for taking photos.
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
972
The tests were performed in a dark room. Four fluorescent light sources were positioned
on the base plane to make the borders of the particles more visible for digital
measurements. The flashlight of the camera was not also utilized during image
acquisition. After the initial projected image (top view image) of the particles is captured
and measured, the photographic stand with camera is rotated 90 degrees so that the
particles are now perpendicular to their original orientation and another projection of the
particles (front view image) is captured and measured (Figure 3b).
(a) (b)
Figure 3: Schematic of the image analysis systems (a) for top view (b) for front view.
Figure 4 shows 3-D view of a regular-shaped solid, a rectangular box. Clearly, both the
top and front views of the solid are identical rectangles. Using a 2-D image analysis setup
consisting of only top camera and capturing an image of each solid would not be
effective in distinguishing the 3-D shapes of the solids (Rao and Tutumluer, 2000).
Figure 4: Three-dimensional views of two regular-shaped solid: rectangular box.
Top View
Front View
S
L
I
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
973
In this research, the particles were placed to the sample tray in their stable position and
two images from top view (longest portion of particle) and front view (thickest portion of
particle) were obtained by imaging system for capturing the each length of particle (i.e.,
small, intermediate, and long) as showed in Figure 5. Images of the top (Figure 5a) and
front views (Figure 5b) were used for the maximum and minimum projected areas of the
particles, respectively. The position of particles was not changed when the two images
were captured.
(a) (b)
Figure 5: (a)Top views of particles, (b) Front views of particles
2.4. Image Processing
Digital-image processing consists of converting camera pictures into digital form and
applying various mathematical procedures to extract relevant information from the
picture. In other words, digital image processing and analysis is concerned with the
transformation and analysis of picture by a computer. The process starts with the capture
of a digital image (usually referred to as imaging) followed by its storage, transmission,
processing, and analysis of the image by computer to extract information or features of
interest (Masad and Sivakumar, 2004).
This section describes the image analysis method used in this study. The output of
camera was a 3872-2592 pixel, 32-bit digital image of RGB color. The particles had to be
identified prior to analysis. ImageJ was used as the image analysis program. Threshold
gray intensity therefore had to be chosen. Thresholding determines the outline of the
aggregate particle in a captured image. Clearly, the particle should have a sharp contrast
against the background to accurately delineate the actual boundaries. A threshold value of
pixel gray level is typically specified, and the actual image is converted to a binary image.
This binary image has only black or white (gray level 0 or 255) pixels to clearly identify
the particle against its background. The gray intensity measured on a given point was
compared to the threshold value. Then, the initial gray image was converted into a binary
image in which the aggregate particles that have lower gray intensity than the threshold
value were set to black while the background was set to white. Applying a global
threshold value for all the image worked well only if the objects of interest (aggregate
particles) had uniform interior gray level and rested upon a background of different, but
uniform, gray level. This was made possible in this study by placing particles on white
background. The original image (32-bit digital image of RGB) (a), 8-bit 256 gray scale
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
974
image (b), 1-bit binary image (c) and output of ImageJ image analysis program (d) are
shown in Figure 6.
Figure 6: Image processing steps
2.5. Grain-Size Distribution with Mechanical and Image Analysis
Mechanical (sieve) analysis depends on measuring mass of particles; whereas image
analysis depends on measuring the number of particles or a single morphological
parameter (i.e., area, length, breadth, equivalent area diameter (dA), mean Feret's
diameter (dF)) of particles. Hence, it is not obvious that the results of these two methods
are comparable (Fernlund et al., 2007). Because image analysis produces area gradation
and uses a particle size definition different from the square sieve size used in mechanical
sieving, direct comparison is not possible, unless the gradation basis and particle size
definitions of the two methods are aligned. For this purpose, a simple method of
converting the area gradation to mass gradation was proposed in literature (Mora et al.,
1998). Moreover, a size correction factor is used to convert the particle sizes measured by
image analysis to equivalent square sieve sizes so that comparison between the image and
mechanical analysis results can be made.
In this research, grain-size distribution of soil samples determined by both mechanical
(sieve) and image analysis methods. In order for the sample to be representative,
approximately 1 kg of each soil sample was taken for mechanical analysis. For image
analysis, the number of particles that can be placed within the measurement area without
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
975
touching or overlapping each other ranges from 15 to about 100, depending on the size
and shape of the particles as shown in Figure 6. Because the soil samples contain more
than 500 particles, all of the particles could not be placed onto the sample tray at same
time. Therefore, the soil samples have to be divided into several sub-samples whose
images are taken successively so that the image of all particles can be captured.
The image processing steps explained above section were applied to all captured images
and the output of image processing program (ImageJ) was obtained. The output of the
ImageJ is area, perimeter, long dimension (length or Feret’s diameter) values in unit of
millimeter, and quantity of particles. These data are transferred to Excel, Microsoft
software. Similar to Lira and Pina (2007), to construct grain-size distribution curves of
the particles, the area gradation was used as described at equation (1). To make
comparison of mechanical and image analysis results, the area gradation points were
transformed to sieve size by extracting the square root of area. Due to the square sieve
openings, this simple method was used. After then, the results are sorted independently
and cumulative curves of the axial relationships plotted. In order to compare results of
top and front views, the images of both views were separately analyzed and plotted.
100*
...100sin 321
AreaTotal
AreaAreaAreaAreaXAreagPasPercent X (1)
2.6. Volume Computation of Particles
Image analysis does not measure mass of particles. However, several researchers have
claimed that more accurate volume and mass determinations are necessary for accurate
construction of grain-size curves (Taylor, 2002; Maerz, 2004; Tutumluer et al., 2005).
Three-dimensional representation of particles is considered very important to be able to
accurately determine the volume (Fernlund et al., 2007). In this sense, a study was
undertaken on gravel soils (S8, S9, and S10). Elongated (S8), flat (S9), and spherical
(S10) particles in gravel soil (8 mm-16 mm) were selected by hand and used for
determination of their volumes. 200 particles were selected for each particle shape. The
particles were also placed three by three within the sample tray for capturing the top and
front images as seen in Figure 5. These images were processed in ImageJ then output data
of ImageJ was transferred to Excel. The output data also contain length (L), breadth (I),
thickness (S), top area and front area of each particle.
Five different methods for calculating volume and hence mass have been tested: Volume
1 is the product of the axial dimensions (cube: V=L*S*I); Volume 2 is the product of the
area of top view and the mean thickness (cylinder1: V= Atop*S); Volume 3 is the product
of the front view area and the longest dimension (cylinder2: V= Afront*L); Volume 4 is
the product of the ellipsoid (V=4/3*Π*((L/2)*(I/2)*(S/2)); Volume 5 is the product of the
modified ellipsoid (V=4/3* Atop*(S/2)). The mass of analyzed particles was determined
with specific gravity. The mass was also calculated by multiplying the specific gravity of
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
976
soil (Gs=2,58) and volume. The grain-size distribution curve was also plotted and
compared with the other results.
3. Results and Discussion
In the following sections, the results of image analysis and mechanical analysis of
particles are presented. The findings of the experimental tests are compared with those of
other studies in the literature and discussed in detail.
3.1. Comparison of mechanical (MA) and image analysis (IA) on GSD
Grain-size distribution of S1 (2mm-4,45mm), S2 (2mm-8mm), S3 (4,75mm-8mm), and
S4 (8mm-16mm) soil samples were determined by using mechanical and image analysis.
The GSD curves are given in Figure 7. It is seen that the GSD curves obtained by IA and
MA do not quite agree with each other, with the curve obtained by MA being consistently
lower than that by IA. The difference between the results of IA and the results of MA
could be attributed that the particle size definitions used in the two analyses are different.
0
20
40
60
80
100
1 10 100Particle Size (mm)
Perc
en
t P
assin
g (
%)
S1-MAS1-IAS2-MAS2-IAS3-MAS3-IAS4-MAS4-IA
Figure 7: GSD curves of four soil samples with different sizes obtained by mechanical
and image analysis (The results of S1, S2, and S3 are also mentioned in Arasan and
Akbulut (2008))
In this study, grain-size distributions (GDS) of soils were estimated by the area
distribution of particles method used by King (1984) and Vallebuona et al. (2003) and
mechanical analysis is also performed by using the mass of particles. Hence, the
difference is obtained from analysis as indicated by Mora et al (1998) and Kwan et al.
(1999). Mora et al. (1998) presented a conversion factor, between 0.81 and 0.89 times the
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
977
length of the intermediated axis and obtained exactly same curves from IA and MA by
using these conversion factors.
On the other hand, some particle properties of soil samples were obtained from the
curves. Table 2 shows these results such as: D10, D30, D50, D60, Cu, Cc, and soil
classification according to the Unified Soil Classification System-USCS. It can be also
seen from Table 2 that D10, D30, D50, and D60 values of IA greater than that MA.
However, the classifications do not change. The soil classification is more important than
GSD curve in soil mechanics. Thus, it could be said that the difference between two
analyses results are insignificant and the size of particles do not significantly affect the
GSD.
Table 2: Some particle properties of soils used in mechanical and image analysis (The
results of S1, S2, and S3 are also mentioned in Arasan and Akbulut (2008))
S1 S2 S3 S4
MA IA MA IA MA IA MA IA
D10 2,3 2,8 2,5 2,9 5,3 6,3 8,8 11,5
D30 2,8 3,3 3,25 3,7 5,9 7,5 9,8 12,9
D50 3,4 4 3,8 4,5 6,3 8,2 11,5 14,1
D60 3,6 4,7 4 5,5 6,6 8,7 12,0 14,5
Cu 1,57 1,68 1,6 1,89 1,25 1,38 1,36 1,26
Cc 0,95 0,83 1,06 0,86 1,00 1,03 0,91 1,00
Soil Classification SP SP SP SP GP GP GP GP
3.2. Influence of volume calculations on GSD
Flat, elongated, and spherical particles in gravel soil (8 mm-16 mm) were selected by
hand as mentioned above and analyzed for determination of their volumes. Five different
methods for calculating volume and hence mass have been tested. The comparison of real
total weight with image analysis results is given in Figure 8. Particle shape did not
significantly affect the weight calculation. However, there are great differences between
the volume calculation methods. It is clearly seen that all of the estimated weights by
volume calculation are higher than real weight and also the modified ellipsoid method
give the best results. The total weights of the S4 (mixed), S8 (elongated), S9 (flat), and
S10 (spherical) were measured manually to be 176gr, 205gr, 126gr, and 143gr,
respectively, while the calculated value by the image analysis approach using a Gs of 2,58
gr/cm3 were 177gr, 212gr, 132gr, and 144gr, respectively for modified ellipsoid method.
These differences in total weights of about 1%, 3%, 5%, and 1% are insignificant.
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
978
0
100
200
300
400
500
Flat Elongated Spherical Mixed
Particle Shape
To
tal
Weig
ht
(gr)
Real WeightModif ied EllipsoidEllipsoidCylinder 1Cylinder 2Cube
Figure 8: The comparison of volume computation type on the total weight of particles
Similarly, Rao et al. (2001) and Tutumluer et al. (2005) proposed a method based on
assuming that the particle is an ellipsoid of revolution for volume calculation. Fernlund et
al. (2007) used a method based on assuming that the particles are ellipsoidal, then the
volume is two-thirds the product of the largest projected area times the thickness.
Furthermore, Rao and Tutumluer (2000) showed that the weights of the aggregates
estimated from the image analysis system agree quite well with the physically measured
values and the errors are clearly biased towards the positive side. They also advocated
that this is expected; as the volumes are slightly overestimated because the cameras
cannot capture information about hollow portions cannot seem by the cameras. The
results of the study described herein demonstrate that the proposed image analysis
method is a promising method for 3D analysis of soil particles.
3.3. Comparison of area (from top and front views) and volume methods
To study the affect of camera positions on the estimated distributions, the results of the
profile area approach applied to two types of analysis (top and front view) were
compared and the results were shown in Figure 9. By comparing the estimated size
distribution when the camera was positioned in the two different positions, it was found
that the top view analysis had the worst results (Figure 9). The reason may due to
direction of particle passing which had the most influence in the image analysis as
indicated previous researchers (Mora et al., 1998; Fernlund, 1998; Kwan et al., 1999).
Fernlund (1998) reported that the least cross-sectional area is most important and usually
the longest dimension of a particle has little effect on sieve results. In this study, the
maximal and least areas of a particle are obtained from top view and front view, as
illustrated in Figure 4 and 5, respectively. In this sense, it could be said that front views
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
979
give more accurate results than top views in image analysis used for grain-size
distribution.
On the other hand, it is clearly seen from Figure 9 difference between the GSD curves of
all methods is insignificant when soil classification results put in consideration. The GSD
curve of soils obtained from volume calculation is similar to the GSD curves of soils
obtained from 2D views (i.e., top and front view) and mechanical analysis. Similarly, Al-
Thyabat et al. (2007) investigated the size distribution of particles moving on a conveyer
belt. They also indicated that profile views give worst results but top and free falling
views are closely to the mechanical views. Consequently, it could be said that using
particle area (2D views) is adequate for obtaining GSD, when contrast using particle
volume/mass (3D view).
0
20
40
60
80
100
1 10 100
Particle Size (mm)
Perc
en
t P
assin
g (
%)
S4-MA
S4-IA(Area-Front View)
S4-IA(Area-Top View)
S4-IA(Volume/Mass-M.Ellipsoid)
Figure 9: The comparison of different views’ image analysis results on GSD curves
3.4. Influence of particle shape on GSD
Particle shape (elongated, flat, and spherical) is an important issue on soil mechanics. For
this reason, effect of particle shape on the GSD was investigated. Size fractions of 4.75
mm-8 mm (S5, S6, S7) and 8 mm-16mm (S8, S9, S10) in soil were selected by hand as
flat, elongated, and spherical and performed the image analysis for grain-size distribution.
Figure 10 shows the grain-size distributions of the samples have 4.75mm-8mm size
fraction. Figure 11 and 12 also show the GSD of samples have 8mm-16mm size fraction
from top and front views, respectively. It is clearly seen that the particle sizes do not
affect the results. The curves from IA in spherical and elongated particles are the nearest
and the farthest to the curve of the mixed particles from MA, respectively (Figure 10, 11).
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
980
Similar to Figure 7, the GSD curves obtained from IA and MA are not compatible with
each others.
0
20
40
60
80
100
1 10 100
Particle Size (mm)
Perc
en
t P
assin
g (
%)
v
MA-Mixed ParticlesIA-Mixed ParticlesIA-Elongated ParticlesIA-Flat ParticlesIA-Spherical Particles
Figure 10: GSD curves of gravel soil (4.75mm-8mm) with different shapes (Arasan and
Akbulut, 2008)
0
20
40
60
80
100
1.0 10.0 100.0
Particle Size (mm)
Perc
en
t P
assin
g (
%)
IA-Flat Particles
IA-Elongated Particles
IA-Spherical Particles
IA-Mixed Particles
MA-Mixed Particles
Figure 11: GSD curves of gravel soil (8 mm-16mm) with different shapes by using top
views of particles
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
981
0
20
40
60
80
100
1.0 10.0 100.0
Particle Size (mm)
Perc
en
t P
assin
g (
%)
IA- Flat Particles
IA-Elongated Particles
IA-Spherical Particles
IA-Mixed Particles
MA-Mixed Particles
Figure 12: GSD curves of gravel soil (8 mm-16mm) with different shapes by using front
views of particles
On the other hand, the curve obtained by MA consistently gave the higher particle size
than that of IA for front views. The GSD curve of spherical and mixed particles obtained
by IA are the nearest curve from the curve of mixed particles obtained by MA. Also, the
GSD curve of flat particles obtained by IA is the farthest curve from the curve of mixed
particles obtained by MA. The difference between the results of top and front views
analysis is explained by the passing mechanism of particles through sieve aperture. As
indicated previously, a particle passes through a sieve by longest dimension. In this sense,
it could be said that the particle form is more effects than particle size to the sieve results.
Similarly, Fernlund (1998) reported that the particle form influences the sieve results.
Taylor (2002) also points out that sieving does not separate particles according to
volume. There is a variation in the volume of particles retained on a sieve due to the
varying shapes of the particles.
It is previously mentioned that the results of sieving are dependent upon the shape of the
particles (Mora et al., 1998; Kwan et al., 1999; Mora and Kwan, 2000; Arasan and
Akbulut, 2008). Similarly, the results of this study showed that the shape of the particles
affected the GSD obtained from image analysis. In this sense, it could be said that the
particle shape is the most important factor on the GSD of soils.
4. Conclusions
The following results are concluded based on the results and on the discussion presented
in this research:
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
982
1. The classification of soils according to the USCS obtained from image and
mechanical (sieve) analysis do not change and the difference between results of
two analyses is insignificant
2. The size of particles does not significantly affect the grain-size distribution.
3. Particle shape does not significantly affect the weight calculation. However, there
are great differences between the volume calculation methods. All of the
estimated weights by volume calculation are higher than real weight and also the
modified ellipsoid method gives the best results.
4. The front views give more accurate results than top views in image analysis used
for grain-size distribution. The difference between the results of top and front
views analysis is explained by the passing mechanism of particles through sieve
aperture.
5. For determination of grain-size distribution, 2D view image analysis is adequate
rather than using volume/mass calculation.
6. The shape of particle significantly affects the grain-size distribution.
In this sense, the image analysis is considered to be a significant technological
advancement towards grain-size distribution of soils with images obtained from digital
cameras from taken photos. It should also be pointed out that further studies on the
determination of index properties of soils by using image analysis are needed to make
more reasonable judgments for utilization of image analysis in geotechnical engineering.
References
1. Alshibli, K.A., Alsaleh, M.I., 2004. Characterizing surface roughness and shape
of sands using digital microscopy. Journal of Computing in Civil Engineering 18
(1), pp 36-45.
2. Al-Thyabat, S., Miles, N.J., Koh, T.S., 2007. Estimation of the size distribution of
particles moving on a conveyor belt. Minerals Engineering 20, pp 72-83.
3. Andriani, G.F., Walsh, N., 2002. Physical properties and textural parameters of
calcarenitic rocks: qualitative and quantitative evaluations. Engineering Geology,
67, pp 5–15.
4. Arasan, S., Akbulut, S., 2008. Determination of grain-size distribution of soils
with image analysis. 12. National Soil Mechanic and Foundation Engineering
Congress, Konya, Turkey. pp. 323-332 (in Turkish with an English summary).
5. Barbery, G., 1991. Mineral Liberation. GB, Canada.
6. Blott, S.J., Pye, K., 2008. Particle shape: a review and new methods of
characterization and classification. Sedimentology 55, pp 31-63.
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
983
7. Bozdag, T., Karpuz, C., 1997. Development of a particle size distribution analysis
system by digital image processing. International Journal of Mining, Reclamation
and Environment 11, pp 59-67.
8. Chandan, C., Sivakumar, K., Masad, E., Fletcher, T., 2004. Application of image
techniques to geometry analysis of aggregate particles. Journal of Computing in
Civil engineering ASCE 18 (1), pp 75-82.
9. Fernlund, J.M.R., 1998. The effect of particle form on sieve analysis: a test by
image analysis. Engineering Geology 50, pp 111–124.
10. Fernlund, J.M.R., 2005a. Image analysis method for determining 3-D shape of
coarse aggregate. Cement and Concrete Research 35, pp 1629-1637.
11. Fernlund, J.M.R., 2005b. 3-D image analysis size and shape method applied to the
evaluation of the Los Angles test. Engineering Geology 77, pp 57-67.
12. Fernlund, J.M.R., 2005c. Image-analysis method for determining 3-D size-
distribution of coarse aggregates. Bulletin of Engineering Geology and the
Environment 64(2), pp 159-166.
13. Fernlund, J.M.R., Zimmerman, R.W., Kragic, D., 2007. Influence of volume/mass
on grain-size curves and conversion of image-analysis size to sieve size.
Engineering Geology 90, pp 124–137.
14. Ghalib, A.M. ve Hryciw, R.D., 1999. Soil particle size distribution by mosaic
imaging and watershed analysis. Journal of Computing in Civil Engineering 13(2),
pp 80-87.
15. King, R.P., 1984. Measurement of particle size distribution by image analyzer.
Powder Technology 39 (2), pp 279–289.
16. Krumbein, W.C.,1941. Measurement and geological significance of shape and
roundness of sedimentary particles. Journal of Sedimentary Petrology 11, pp 64–
72.
17. Kuo, C.Y., Frost, J.D., Lai, J.S., Wang, L.B., 1996. Three-Dimensional Image
Analysis of Aggregate Particles from Orthogonal Projections. Transportation
Research Record 1526, National Research Council, Washington D.C. pp. pp 98-
103.
18. Kwan, A.K.H., Mora, C.F., Chan, H.C., 1999. Particle shape analysis of coarse
aggregate using digital image processing. Cement and Concrete Research 29, pp
1403–1410.
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
984
19. Lanaro, F., Tolppanen, P., 2002. 3D characterization of coarse aggregates.
Engineering Geology 65, pp 17-30.
20. Lee, J.R.J., Smith, M.L., Smith, L.N., 2007. A new approach to the three-
dimensional quantification of angularity using image analysis of the size and form
of coarse aggregates. Engineering Geology 91, pp 254:264.
21. Lin, C.L., Miller, J.D., Herbst, J.A., 1987. Solutions to the transformation
equations for volumetric grade distribution from linear and/or areal grade
distributions. Powder Technology 50, pp 55–63.
22. Lira, C., Pina, P., 2007. Sedimentological analysis of sands. IbPRIA 2007. Part II.
LNCS 4478, pp. 388-395.
23. Maerz, N.H., 2004. Technical and Computational aspects of the measurement of
aggregate shape by digital image analysis. Journal of Computing in Civil
engineering ASCE 18 (1), pp 10-18.
24. Maerz, N.H., Lusher, M., 2001. Measurement of flat and elongation of coarse
aggregate using digital image processing. Paper No. 01- 0177, Transportation
Research Board 80th Annual Meeting, 14 pp.
25. Masad, E., Saadeh, S., Rousan, T.A., Garboczi, E., Little, D., 2005. Computations
of particle surface characteristics using optical and X-ray CT images.
Computational Materials Science 34, pp 406-424.
26. Masad, E., Sivakumar, K., 2004. Advanced in the characterization and modeling
of civil engineering materials using imaging techniques. Journal of Computing in
Civil Engineering 18(1), pp 1-1.
27. Mertens, G., Elsen, J., 2006. Use of computer assisted image analysis for the
determination of the grain-size distribution of sands used in mortars. Cement and
Concrete Research 36, pp 1453-1459.
28. Mora C.F., Kwan A.K.H., 2000. Sphericity, Shape Factor, and Convexity
Measurement of Coarse Aggregate for Concrete Using Digital Image Processing.
Cement and Concrete Research 30 (3), pp 351-358.
29. Mora, C.F., Kwan, A.K.H., Chan, H.C., 1998. Particle Size Distribution Analysis
Of Coarse Aggregate Using Digital Image Processing. Cement and Concrete
Research. 28(6), pp 921–932.
30. Rao, C., Tutumluer, E., 2000. Determination of volume of aggregates: New
image-analysis approach. Transportation Research Record 1721, Transportation
Research Board, National Research Council, Washington D.C., pp 73–80.
INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING
Volume 1, No 4, 2011
© Copyright 2010 All rights reserved Integrated Publishing services
Research article ISSN 0976 – 4399
985
31. Rao, C., Tutumluer, E., Stefanski, J.A., 2001. Coarse aggregate shape and size
properties using a new image analyzer. ASTM Journal of Testing and Evaluation
29, pp 79–89.
32. Sneed, E.D. and Folk, R.L., 1958. Pebbles in the Lower Colorado River, Texas: a
study in particle morphogenesis. Journal of Geology 66, pp 114–150.
33. Taylor, M.A., 2002. Quantitative measures for shape and size of particles. Powder
Technology 124, pp 94–100.
34. Tolppanen, P., Stephansson, O., Stenlid, L., 2002. 3-D degradation analysis of
railroad ballast. Bulletin of Engineering Geology and the Environment 61, pp 35-
42.
35. Tutumluer, E., Pan, T., Carpenter, S.H., 2005. Investigation of aggregate shape
effects on hot mix performance using an image analysis approach. Civil
Engineering Studies. Transportation Engineering Series 137 University Illinois,
Urbana-Champaign, 122 pp.
36. Vallebuona, G., Arbro, K., Casali, A., 2003. A procedure to estimate particle
distributions from area measurements. Minerals Engineering 16, pp 323–329.
37. Zingg, T., 1935. Bietrahe zur Schotteranalyse. Schweiz Mineral Petrography 15,
pp 139-140.