AWERProcedia
Information Technology
&
Computer Science
1 (2012) 24-29
2nd World Conference on Information Technology (WCIT-2011)
Malaysian wood identifier using image histogram variations
K. Jeyaprakash A/L Kunjambua, Jasni Mohamad Zaina*
a
Faculty of Computer System and Software Engineering, University Malaysia Pahang, Malaysia
Abstract
Wood identification is very important to enable the wood can be used for its correct usage. Unfortunately some of the
organizations in Malaysia are still using traditional method to identify the wood type which can be influenced by human
error. Image processing technology can be used to solve this problem. The image of the wood block sample will be used as
the input to identify the type of the wood. This paper investigates whether the grayscale image pixel value variations in the
image histogram can be used to differentiate the wood images. First the wood sample image will be converted to grayscale
then reagent of the interested will be extracted from the converted image. Then this will be used to create histogram. Data
from the histogram will be used to produce a set of data that contains all the pixel values that present in the image. Then this
data set will be compared with the Tested Sample Set data to get the differences. The results proved that the proposed
method can be used for wood identification
Keywords : Wood identification; Image processing; Image histogram
Selection and/or peer review under responsibility of Prof. Dr. Hafize Keser.
©
2 Academic World Education & Research Center. All rights reserved.
1. Introduction
Every type of wood has different strength, durability and density. This makes the industrial price of the wood
is different from one another [1]. It is important to differentiate these woods correctly to use them for their
appropriate uses. This is because wrongly identified wood could cause huge impact [2]. Because, some woods
are being used as important structure of buildings in the construction field. Unfortunately some Malaysian
organizations are still using traditional methods to identify the wood species. But this result can be influenced by
the human error.
Wood identification process requires a lot of knowledge and experience to name the wood based on its
* ADDRESS FOR CORRESPONDENCE: K.Jeyaprakash A/L Kunjambu. Tel.: +609 5492133.
E-mail address: jasni@ump.edu.my.
25
K.Jeyaprakash A/L Kunjambu/ AWERProcedia Information Technology & Computer Science (2012) 24-29
structure. Because the identification requires careful observation and every single things that can be found on
the cross section must be observed and considered for the accurate identification [5]. Even a single mistake
could give the wrong result. Moreover, wood identification by human eyes often leads to error due to visual
stress, and tiredness. Therefore, wood identification process result that was carried out by human tent to be not
accurate due to the human errors.
Other than that, identifying a wood is very long procedures where, for some unusual woods we have do a lot
references. So it will take a very long to identify the wood. Other than that, some woods needed to be tested in
lab to identify it. Normally, this identification method takes 2-7 days to be completed [2]. So this will further
increase the identification period. Wood identification also involves a lot of knowledge [6] and experience and a
very careful observational skill, but there are only few people that could identify the structure accurately. So this
creates a shortage in experts in this field.
2. Method
In this section, we will discuss about the current method that being used by MTIB to the wood identification
and the proposed method [2].
2.1. Current method
There are two methods that are being used by MTIB. They are wood identification based on physical
structures and identification based on structural features.
2.1.1. Identification Based on physical features
Indentifying the wood based on the physical characteristic is the preliminary way of timber identification. The
wood is identified by the observation on certain physical characteristic such as color, pattern, texture, hardness,
weight and odor of the wood sample. It is not necessary to have any kind of equipments to identify the wood
sample based on this method. But, this method is only can be done by someone who has the great observation
skill and someone who is well experienced in the field. It only takes less than 5 minutes to identify the wood
using this method. Even though, it is easier way to identify the wood through this method the result is not
consistent and sometime can be affected by the human error. And only certain species are able to be identified
using this method.
2.1.2. Identification Based on structural features
Wood identification using this method needs the usage of equipments such as hand lens and microscope.
These equipments enable the clear view of the wood structure. Hand lens can be used to identify all the
hardwood that can be found at Malaysia. It only takes less than 15 minutes to identify the wood using this hand
lens. Before the observation can done the wood have to be sliced using the pen knife to get the clear picture of
the wood sample [3]. Using hand lens way to identify the wood sample takes great observational skill and also a
well experienced person to give the correct result [7]. Although, this method gives consistent result, it can be
affected by the human error [2].
Unlike the previous method, using microscopic way to identify the wood sample require the preparation of
sample in the form of microscope specimen slide or wood block. The microtome is used to cut very thin slices of
a specimen in order examine the specimen. Although, the microscopic method gives the most accurate result,
the wood specimen preparation is the very complicated process to be done. Normally it will take about two to
seven days to identify the wood using the microscopic way [2]. This is because the wood samples have to be
prepared in the chemical lab using the chemical solution so that it can be observed under the microscope.
26
K.Jeyaprakash A/L Kunjambu/ AWERProcedia Information Technology & Computer Science (2012) 24-29
2.1.3 Proposed method
In this work, the image processing technology will be used in the wood identification method. The image
histograms will be used to differentiate the images. For this purpose three species will be used as the samples to
investigate whether the proposed method can be used to identify these three species. The species that was
selected are durian, ramin and rengas. The wood images are shown in the figure below (figure 1).
Figure 1: (a)
(b)
(c)
Figure 1 is showing the grayscale images of the sample species. (a) image of durian wood, (b) image of rengas,
(c) image of ramin wood
First of all, the sample images will be converted to grayscale images. Next, these images will go through image
segmentation process. Where only some required part of the image will be used further processes. This will
create the reagent of interest. This reagent of interest will be used to create image histograms.
2.2.1. Image Histogram
Image histogram is a graphical representation of the image pixel variation. It shows us the current tonal range
of our image [4]. Histograms work equally well with full color images and also gray scale images. The horizontal
axis of the histogram will represent the pixel range of the image. For a grayscale image the range of horizontal
image will be 0 to 255. On the other hand, the vertical axis will represent the frequency of the pixel value.
Figure 2: (a)
(b)
(c)
Figure 2 is showing the grayscale histogram images of the sample species. (a) image histogram of durian wood, (b) image
histogram of rengas, (c) image histogram of ramin wood
From the histograms above (figure 2) we can see the variations of the pixel values that are differ for each
species. This difference can be used to differentiate the species as well as can be used to identify them. To do
that the pixel values that present in the images have to be identified.
27
K.Jeyaprakash A/L Kunjambu/ AWERProcedia Information Technology & Computer Science (2012) 24-29
Table 1. The first five pixels of each species
Pixel Position
Durian
Ramin
Rengas
1 Pixel
0
12
26
2nd Pixel
17
14
38
3 Pixel
26
15
45
4th Pixel
36
16
46
38
17
47
st
rd
th
5 Pixel
From table1 we can see the variation of the pixel value among the species. The pixel value of the durian wood
image is started with zero and the pixel values are not closely related. But the pixel values of ramin wood are
closely related. The differences among the pixel values are small. On the other hand, the pixel value of rengas
has started with larger number compared to durian wood image and they does not closely related.
These pixel variations can be used to differentiate the wood species. But before that the pixel value
distribution patterns have to be determined in order to use this method.
2.2.2. Pixel Averaging
In order to detect the pixel value distribution pattern, ten images from each wood species will be used to
determine the average pixel value of each pixel range (from 0 to 255). When the pixel value presents in the
image it will be noted as the value of the pixel and when the pixel value is absent in the image it will be noted as
ze o
.
Table 2. The average pixel value of ramin wood pixel range
Pixel range value ( i )
Ramin1
Ramin2
...
Ramin10
Average (āi)
0
0
0
…
0
0
1
0
0
…
1
0.3
2
0
0
…
2
0.4
3
0
3
…
3
1.2
4
0
4
…
4
1.6
…
…
…
…
…
…
255
255
0
…
255
255
The table above (table 2) is showing the average pixel value of ramin wood. These values are called Tested
Sample Set and noted with āi sign. The i is the pixel a ge alue of the a e age alue. Tested Sample Set value
ill e al ulated fo all th ee ood spi es. O e this p o ess is do e, e a use the ā i value to identify the
wood images.
2.2.3 Image Recognition
Once the Tested Sample Set has been determined this value can be used to identify the image. The image that
needs to be identified will be converted to grayscale image and the grayscale image will be used to create image
histogram. This followed by calculation of pixel values from image histogram. The calculated value will be notes
as Xi. The i alue ill e a ged f o
to 55. This alue ill e used i i age e og itio p o ess as follo s
the equation below.
∑ (i=0 to i=255) = āi - Xi)²
(1)
28
K.Jeyaprakash A/L Kunjambu/ AWERProcedia Information Technology & Computer Science (2012) 24-29
The pixel values that will be gained from the image that needed to be identified will be compared with all three
Tested Sample Sets as follows the equation1.
3. Results
To investigate the effectiveness of the proposed method, a random image of the durian was selected to test
using the proposed method. The image was converted to grayscale and undergoes image segmentation process.
Then the reagent of interest was used to create image histogram and the image pixel range was calculated from
the histogram. Then these values were compared with the Tested Sample Sets using the equation1. Other than
that before the result will be displayed the system will compare the answer with the possible range of the
answer which already will be calculated from during data averaging. When the answer is beyond or lesser than
the range value of the particular species it will be declared as the image is not durian, rengas and ramin. The
results are shown in the table 3 below.
Table3. The square root of equation 1 result
Wood species
E uatio
Durian
291
Rengas
595
Ramin
2163
esult ½
From the above table (table 3) we can see that the comparison of the random image with durian Tested Sample
Set values is showing the lowest number. So this means the random image is the image of durian wood.
4. Discussion
The results from the table above proved that the proposed method can be used for wood identification. This
method can be a solution for the problems that are being faced by Malaysian timber industries. Since this
system will be using computer it would not affected by the human error. Other than that, anybody from the
timber industries can use this system to identify the wood type. This system also has some limitations, where the
image acquisition process of this method must be conducted in the fixed environment. This means, the things
such as the light intensity, the camera, distance of wood block from camera lens must be constant. So, all the
images that will be used by the method must be captured under the same conditions. And the size of reagent of
interest must be constant as well.
5. Conclusion
Wood identification is the important process in the timber related industries. They are many methods are
being used by timber industries for wood identification purpose. This paper proposes a method that can be used
to identify the wood images. This method follows few stages to identify the wood. First the wood sample image
will be converted to grayscale then reagent of the interested will be extracted from the converted image. Then
this will be used to create histogram. Data from the histogram will be used to produce a set of data that contains
all the pixel values that present in the image. Then this data set will be compared with the Tested Sample Set
data to get the differences using the equation1.
However the users have to follow the restrictions that have been stated in the discussion section to get reliable
result.
29
K.Jeyaprakash A/L Kunjambu/ AWERProcedia Information Technology & Computer Science (2012) 24-29
References
[1] Menon, P. (1993). Structure and Identification of Malaysain Woods. Kepong: Malaysian Forest Records.
[2] Muhammad, K. B. (2010). WOOD IDENTIFICATION USING IMAGE PROCESSING. Cheras: Malaysian Timber Industrial Board(MITB).
[3] Issabelle Guyon, S. G. (2006). Features extraction Foundation and Applications.
[4] Woods, R. G. (1992). Digital Image Processing. Addison-Wesley Publishing Company.
[5] Forest Research Institute Malaysia. (2009, December 9). e-woodID. Retrieved May 4, 2011, from FRIM: info.frim.gov.my/woodid/ewwodID-introduction7Dec09.pd
[6] Elisabeth A. Wheeler, P. B. (1998). Wood Identificstion - A Review. Wood Identificstion - A Review , 24
[7] Society of Wood Science and Technology. (2002). Wood Identification: Equipment & Technique. Madison: Society of Wood Science and
Technology.