International Journal of Computer Applications (0975 – 8887)

Volume 17– No.1, March 2011

Fast and Accurate Detection and Classification of

Plant Diseases
H. Al-Hiary, S. Bani-Ahmad, M. Reyalat, M. Braik and Z. ALRahamneh
Department of Information Technology
Al-Balqa’ Applied University, Salt Campus, Jordan

ABSTRACT plant diseases is expensive, inefficient, and difficult. Also, it

requires the expertise of trained botanist [1].
We propose and experimentally evaluate a software solution
for automatic detection and classification of plant leaf diseases. In [4] the authors have worked on the development of methods
The proposed solution is an improvement to the solution for the automatic classification of leaf diseases based on high
proposed in [1] as it provides faster and more accurate solution. resolution multispectral and stereo images. Leaves of sugar
The developed processing scheme consists of four main phases beet are used for evaluating their approach. Sugar beet leaves
as in [1]. The following two steps are added successively after might be infected by several diseases, such as rusts (Uromyces
the segmentation phase. In the first step we identify the mostly- betae), powdery mildew (Erysiphe betae).
green colored pixels. Next, these pixels are masked based on In [20], a fast and accurate new method is developed based on
specific threshold values that are computed using Otsu's computer image processing for grading of plant diseases. For
method, then those mostly green pixels are masked. The other that, leaf region was segmented by using Otsu method [15; 12;
additional step is that the pixels with zeros red, green and blue 13]. After that the disease spot regions were segmented by
values and the pixels on the boundaries of the infected cluster using Sobel operator to detect the disease spot edges. Finally,
(object) were completely removed. The experimental results plant diseases are graded by calculating the quotient of disease
demonstrate that the proposed technique is a robust technique spot and leaf areas.
for the detection of plant leaves diseases. The developed
Studies show that Machine learning methods can successfully
algorithm‟s efficiency can successfully detect and classify the
be applied as an efficacious disease detection mechanism.
examined diseases with a precision between 83% and 94%, and
Examples of such machine learning methods that have been
can achieve 20% speedup over the approach proposed in [1].
applied in agricultural researches; Artificial Neural Networks
General Terms (ANNs), Decision Trees, K-means, k nearest neighbors, and
Support Vector Machines (SVMs). For example, Wang et al. in
Artificial Intelligence, Image Processing. [19] predicted Phytophthora infestans infection on tomatoes by
using ANNs. Also, Camargo and Smith in [5] used SVMs to
Keywords identify visual symptoms of cotton diseases using SVMs.
K-means, SGDM Matrix, Color Co-occurrence Method, HSI,
Table 1: Summary of total losses due to disease damage and
Neural Networks.
cost of control in Georgia, USA in 2007 [11].
Crop Value of damage Cost of
1. INTRODUCTION ($ millions) control
Plant diseases have turned into a dilemma as it can cause Apple 0.073 0.267
significant reduction in both quality and quantity of Blueberry 0.14 2.555
agricultural products [20]. It is estimated that 2007 plant Bunch Grape 0.112 0.27
disease losses in Georgia (USA) is approximately $539.74 Corn 12.4 0.5
million. Of this amount, around 185 million USD was spent on Cotton 81.7 12.2
controlling the diseases, and the rest is the value of damage
Muscadine Grape 0.026 0.096
caused by the diseases. Those numbers are listed in Table 1.
Ornamental 41.22 21.2
The naked eye observation of experts is the main approach Peach 0.177 3.19
adopted in practice for detection and identification of plant Peanut 58.7 41.2
diseases [20]. However, this requires continuous monitoring of Pecan 0.64 17.4
experts which might be prohibitively expensive in large farms. Soybean 5.3 1.9
Further, in some developing countries, farmers may have to go Strawberry 0.32 0.683
long distances to contact experts, this makes consulting experts Turf 126.6 61.2
too expensive and time consuming [14; 5; 8].
Vegetable 18.1 20.6
Automatic detection of plant diseases is an essential research Wheat 0.99 1.8
topic as it may prove benefits in monitoring large fields of Totals 346.49 185.06
crops, and thus automatically detect the symptoms of diseases
as soon as they appear on plant leaves [1; 18; 8]. Therefore;
looking for fast, automatic, less expensive and accurate method Three are two main characteristics of plant-disease detection
to detect plant disease cases is of great realistic significance machine-learning methods that must be achieved, they are:
[14; 5]. speed and accuracy.
Machine learning based detection and recognition of plant In this study an automatic detection and classification of leaf
diseases can provide clues to identify and treat the diseases in diseases has been proposed, this method is based on K-means
its early stages [8; 18]. Comparatively, visually identifying as a clustering procedure and ANNs as a classifier tool using

31
 International Journal of Computer Applications (0975 – 8887)
Volume 17– No.1, March 2011

some texture feature set. The proposed method is an Algorithm 1: Basic steps describing the proposed algorithm.
improvement of the approach proposed in [1].
As a testbed we use a set of leaves which are taken from Al- 1 RGB image acquisition
Ghor area in Jordan. Our program has been tested on five 2 Create the color transformation structure
diseases which affect on the plants; they are: Early scorch,
3 Convert the color values in RGB to the space specified in
Cottony mold, Ashen mold, late scorch, and tiny whiteness.
the color transformation structure
The proposed framework could successfully detect and classify
the tested diseases with precision of more than 94% in average 4 Apply K-means clustering
with more than 20% speedup over the presented approach in 5 Masking green-pixels
[1]. The minimum precision value was 84% compared to 80%
precision in the previous approach. 6 Remove the masked cells inside the boundaries of the
infected clusters
In conclusion, the aim of this work is threefold: 1) identifying
the infected object(s) based upon K-means clustering 7 Convert the infected (cluster / clusters) form RGB to HSI
procedure; 2) extracting the features set of the infected objects Translation
using color co-occurrence methodology for texture analysis; 3) 8 SGDM Matrix Generation for H and S
detecting and classifying the type of disease using ANNs, 9 Calling the GLCM function to calculate the features
moreover, the presented scheme classifies the plant leaves and
stems at hand into infected and not-infected classes. 10 Texture Statistics Computation
11 Configuring Neural Networks for Recognition
2. THE PROPOSED APPROACH – STEP-
BY-STEP DETAILS
The overall concept that is the framework for any vision
related algorithm of image classification is almost the same.
First, the digital images are acquired from the environment
using a digital camera. Then image-processing techniques are
applied to the acquired images to extract useful features that
are necessary for further analysis. After that, several analytical
discriminating techniques are used to classify the images
according to the specific problem at hand. Figure 1 depicts the
Early scorch Cottony mold
basic procedure of the proposed vision-based detection
algorithm in this research.

Image acquisition

Image preprocessing

Image segmentation
Ashen mold Late scorch
Feature extraction
Statistical analysis Figure 2: Sample images from our dataset.
Classification based on a
classifier

Figure 1: The basic procedure of the proposed image-

processing-based disease detection solution
The proposed approach step - by - step of the image
segmentation and recognition processes is illustrated in
Algorithm 1. In the initial step, the RGB images of all the leaf
samples were picked up. Some real samples of those diseases Figure 3: A leaf image infected with tiny whiteness
are shown in Figure 2. It is obvious from Figure 2 that leaves disease (left) and a normal leaf image (right).
belonging to early scorch, cottony mold, ashen mold and late
scorch have significant differences form greasy spot leaves in Figure 4 shows an example of a plant stem that is infected with
terms of color and texture. Also, Figure 3 shows two images; white mold disease.
the left image is infected with tiny whiteness disease, and the
right image is a normal image. However, the leaves related to
these six classes (early scorch, cottony mold, ashen mold, late
scorch, tiny whiteness and normal) had very small differences
as discernible to the human eye, which may justify the
misclassifications based on naked eye.

32
 International Journal of Computer Applications (0975 – 8887)
Volume 17– No.1, March 2011

of the device used to draw it. Thus, we created the color

transformation structure that defines the color space
conversion. Then, we applied the device-independent color
space transformation, which converts the color values in the
image to the color space specified in the color transformation
structure. The color transformation structure specifies various
parameters of the transformation. A device dependent color
space is the one where the resultant color depends on the
equipment used to produce it. For example the color produced
using pixel with a given RGB values will be altered as the
Figure 4: A plant stem that is infected with the white mold disease brightness and contrast on the display device used. Thus the
RGB system is a color space that is dependent.
In details, in step 2 a color transformation structure for the The K-means clustering algorithm tries to classify objects
RGB leaf image is created, and then, a device-independent (pixels in our case) based on a set of features into K number of
color space transformation for the color transformation classes. The classification is done by minimizing the sum of
structure is applied in step 3. Steps 2 and 3 are inevitable for squares of distances between the objects and the corresponding
carrying out step 4. In this step the images at hand are cluster or class centroid [10; 7]. However, K-means clustering
segmented using the K-Means clustering technique [10; 7; 3; is used to partition the leaf image into four clusters in which
9]. These four steps constitute phase 1 whereas, the infected one or more clusters contain the disease in case when the leaf
object (s) is/are determined. In step 5, we identify the mostly- is infected by more than one disease. In our experiments
green colored pixels. After that, based on specified and varying multiple values of number of clusters have been tested. Best
threshold value that is computed for these pixels using Otsu's results were observed when the number of clusters was 3 or 4.
method [12; 13], these mostly green pixels are masked as A stem image infected with early scorch and its first cluster
follows: if the green component of pixel intensities is less than (the infected object itself) is shown in Figure 5.
the pre-computed threshold value, the red, green and blue
components of the this pixel is assigned to a value of zero. This
is done in sense that these pixels have no valuable weight to
the disease identification and classification steps, and most
probably those pixels represent healthy areas in the leave.
Furthermore, the image processing time should become
significantly reduced.
In step 6 the pixels with zeros red, green and blue values and
the pixels on the boundaries of the infected cluster (object)
were completely removed. Steps 5 and 6 form phase 2, and this a) b)
phase is helpful as it gives more accurate disease classification
Figure 5: A stem image infected with early scorch; a) original
and identification results with satisfied performance and the
image b) cluster 1 image
overall computation time should become significantly less. The
observations behind steps 5 and 6 were experimentally An example of the output of K-Means clustering for a leaf
validated. infected with early scorch disease is shown in Figure 6. It is
observed from Figure 6 that cluster 4 contains infected object
Next, in step 7 the infected cluster was then converted from
of early scorch disease. Furthermore, clusters 1 and 2 contain
RGB format to HSI format. In the next step, the SGDM
the intact parts of leaf, although they are distinct from each
matrices were then generated for each pixel map of the image
other. However, cluster 3 represents the black background of
for only H and S images. The SGDM is a measure of the
the leaf which can be discarded primarily. Finally, the image in
probability that a given pixel at one particular gray-level will
(e) facilitates the segmentation procedure followed in K-means
occur at a distinct distance and orientation angle from another
algorithm.
pixel, given that pixel has a second particular gray-level. From
the SGDM matrices, the texture statistics for each image were 2.2 Phase 2 – Masking the green pixels and
generated. Concisely, the features set were computed only to
pixels inside the boundary of the infected areas of the leaf. In the pixels on the boundaries
other words, healthy areas inside the infected areas were also This phase consists of two steps: The mostly green colored
removed. Steps 7 – 10 form phase 3 in which the texture pixels are identified, and then the global image threshold using
features for the segmented infected objects in this phase are Otsu's method [12; 13] has been applied in order to specify the
calculated. Finally, the recognition process in the fourth phase varying threshold value which chooses the threshold to
was performed to the extracted features through a pre-trained minimize the intraclass variance of the thresholded black and
neural network. For each image in the data set the subsequent white pixels. Next, the green pixels are masked as follows: if
steps in Algorithm 1 were repeated. the green component of pixel intensities is less than the
The Proposed approach for segmentation and classification computed threshold value, then, the red, green and blue
plant diseases can be divided into four phases: components of the this pixel are cleared. The next step in this
phase is focused on deleting both the pixels with zeros
2.1 Phase 1 - K-means Clustering Technique components and the pixels on the boundaries of the infected
cluster(s).
There are two preprocessing steps that are needed in order to
implement the K-means clustering algorithm: The phase starts 2.3 Phase 3 – Features Extraction
first by creating device-independent color space transformation
structure. In a device independent color space, the coordinates In the proposed approach, the method adopted for extracting
used to specify the color will produce the same color regardless the feature set is called the Color Co-occurrence Method or

33
 International Journal of Computer Applications (0975 – 8887)
Volume 17– No.1, March 2011

CCM method in short. It is a method, in which both the color Color spaces can be transformed from one space to another
and texture of an image are taken into account, to arrive at easily. In our experiments, the Equations 1, 2 and 3 were used
unique features, which represent that image. to transform the image‟s components from RGB to HSI [6]:

(2)

(3)

(a) (b) The color co-occurrence texture analysis method was

developed through the use of Spatial Gray-level Dependence
Matrices (SGDM‟s) [20]. The gray level co-occurrence
methodology is a statistical way to describe shape by
statistically sampling the way certain grey-levels occur in
relation to other grey-levels.
These matrices measure the probability that a pixel at one
particular gray level will occur at a distinct distance and
orientation from any pixel given that pixel has a second
particular gray level. For a position operator p, we can define a
matrix Pij that counts the number of times a pixel with grey-
(c) (d) level i occurs at position p from a pixel with grey-level j. The
SGDMs are represented by the function P(i, j, d, Ө) where i
represents the gray level of the location (x, y) in the image I(x,
y), and j represents the gray level of the pixel at a distance d
from location (x, y) at an orientation angle of Ө. The reference
pixel at image position (x, y) is shown as an asterix. All the
neighbors from 1 to 8 are numbered in a clockwise direction.
Neighbors 1 and 5 are located on the same plane at a distance
of 1 and an orientation of 0 degrees. An example image matrix
and its SGDM are already given in the three equations above.
In this research, a one pixel offset distance and a zero degree
(e) (f) orientation angle was used.
Figure 6: An example of the output of K-Means clustering for a After the transformation processes, we calculated the feature
leaf that is infected with early scorch disease. (a) The infected set for H and S, we dropped (I) since it does not give extra
leaf picture. (b, c, d, e) the pixels of the first, second, the third information. However, we use GLCM function in Matlab to
and the fourth clusters. (e) a single gray-scale image with the create gray-level co-occurrence matrix; the number of gray
pixels colored based on their cluster index. levels is set to 8, and the symmetric value is set to “true”, and
finally, offset is given a” 0” value.

2.3.1 Co-occurrence Methodology for Texture Analysis 2.3.2 Normalizing the CCM matrices
The image analysis technique selected for this study was the The CCM matrices are then normalized using Equation 4.
CCM method. The use of color image features in the visible
light spectrum provides additional image characteristic features
over the traditional gray-scale representation [2]. (4)
The CCM methodology established in this work consists of is the image attribute matrix, represents the
three major mathematical processes. First, the RGB images of intensity co-occurrence matrix, and represents the total
leaves are converted into Hue Saturation Intensity (HSI) color number of intensity levels.
space representation. Once this process is completed, each
pixel map is used to generate a color co-occurrence matrix, The marginal probability matrix ( ) can be defined as shown
resulting in three CCM matrices, one for each of the H, S and I in Equation 5.
pixel maps. (HSI) space is also a popular color space because it (5)
is based on human color perception [17]. Electromagnetic
radiation in the range of wavelengths of about 400 to 700 Sum and difference matrices ( ) are defined as shown
nanometers is called visible light because the human visual in Equations 6 and 7, respectively.
system is sensitive to this range. Hue is generally related to the
wavelength of a light and intensity shows the amplitude of a (6)
light. Lastly, saturation is a component that measures the
“colorfulness” in HSI space [17].
Where ,

34
 International Journal of Computer Applications (0975 – 8887)
Volume 17– No.1, March 2011

(7) Before the data can be fed to the ANN model, the proper
network design must be set up, including type of the network
Where . and method of training. This was followed by the optimal
parameter selection phase. However, this phase was carried out
2.3 Texture Features Identification simultaneously with the network training phase, in which the
The following features set were computed for the components network was trained using the feed-forward back propagation
H and S: network. In the training phase, connection weights were always
updated until they reached the defined iteration number or
acceptable error. Hence, the capability of ANN model to
The angular moment ( ) is used to measure the respond accurately was assured using the Mean Square Error
homogeneity of the image, and is defined as shown in (MSE) criterion to emphasis the model validity between the
Equation 8. target and the network output.

(8)
3. EXPERIMENTAL RESULTS AND
OBSERVATIONS
The produc moment (cov) is analogous to the
covariance of the intensity co-occurance matrix and 3.1 Input Data Preparation and
is defined as shown in Equation 9. Experimental Settings
(9) In our experiments, two main files were generated, namely: (i)
Training texture feature data, and (ii) Testing texture feature
data. The two files had 192 rows each, representing 32 samples
The sum and difference entropies ( and ) which from each of the six classes of leaves. Each row had 10
are computed using Equations 10 and 11 columns representing the 10 texture features extracted for a
respectively. particular sample image. Each row had a unique number (1, 2,
3, 4, 5 or 6) which represented the class (i.e., the disease) of the
(10) particular row of data. „1‟ represented early scorch disease
infected leaf. „2‟ represented Cottony mold disease infected
(11) leaf. „3‟ represented ashen mold disease infected leaf. „4‟
represented late scorch disease infected leaf. „5‟ represented
tiny whiteness disease infected leaf, and „6‟ represented normal
The entropy feature ( ) is a measure of the amount of leaf. Then, a software program was written in MATLAB that
order in an image, and is computed as as defined in would take in .mat files representing the training and testing
Equation 12. data, train the classifier using the „train files‟, and then use the
(12) „test file‟ to perform the classification task on the test data.
Consequently, a Matlab routine would load all the data files
The information measures of correlation ( ). (training and testing data files) and make modifications to the
is defined as shown in Equation 13. data according to the proposed model chosen. In the
experimental results, the threshold value for each of the above
(13) categories is constant for all samples infected with the same
Where: disease. This threshold is a global image threshold that is
computed using Otsu's method [12; 13].
The architecture of the network used in this study was as
Contrast ( ) of an image can be measured by the follows. A set of 10 hidden layers in the neural network was
inverse difference moment as shown in Equation 14. used with the number of inputs to the neural network (i.e. the
number of neurons) is equal to the number of texture features
(14) listed above. The number of output is 6 which is the number of
classes representing the 5 diseases studied along with the case
of normal (uninfected) leaf. Those diseases are early scorch,
Correlation ( is a measure of intensity linear cottony mold, ashen mold, late scorch, tiny whiteness. The
dependence in the image and is defined as shown in neural network used is the feed forward back propagation with
Equation 15. the performance function being the Mean Square Error (MSE)
and the number of iterations was 10000 and the maximum
(15) allowed error was 10-5.

3.2 Experimental Results

2.4 Phase 4 – Neural Networks
In this paper, neural networks are used in the automatic The results for NN classification strategy for testing samples
detection of leaves diseases. Neural network is chosen as a were given in Table 2. These results were obtained using a NN
classification tool due to its well known technique as a classifier for different five diseases. In particular, model M1
successful classifier for many real applications. The training achieved the highest overall classification accuracy, in which it
and validation processes are among the important steps in achieved an overall accuracy of 94% compared to the 89.5%
developing an accurate process model using NNs. The dataset accuracy achieved in [1]. Also, it can be concluded that model
for training and validation processes consists of two parts; the M1 is the best overall model in this classifier in terms of
training feature set which are used to train the NN model; accuracy and in computational time for both training and
whilst a testing features sets are used to verify the accuracy of classification. In conclusion, Table 2 reported better
the trained NN model. classification accuracies for all the data models. Also, Figure 7

35
 International Journal of Computer Applications (0975 – 8887)
Volume 17– No.1, March 2011

shows a graph that representing the percentage classification of Table 4: Classification results per class for neural network with
various diseases of all the data models shown in Table 2. back propagation.
Table 2: Percentage classification of various diseases

Late scorch

whiteness

Accuracy
Cottony

Normal
species

scorch

Ashen
From

Early
mold

mold

Tiny
Early scorch

Ashen mold

Late scorch

whiteness
Features

Cottony

average
Normal

Overall
Model

Color

mold

Tiny
used

Early 25 0 0 0 0 1 100
M1 HS 98 96 89 91 92 100 94.33 scorch
Cottony 0 24 0 1 0 0 96
M2 H 90 92 86 89 93 98 91.33 mold
M3 S 90 89 85 89 81 98 88.67 Ashen 0 0 25 0 0 1 100
M4 I 92 89 84 88 86 99 89.67 mold
Late 0 0 0 22 1 0 88
M5 HSI 81 84 78 79 81 99 83.67 scorch
Tiny 0 1 0 0 23 0 92
whiteness
120 Normal 0 0 0 2 1 23 92
100 Early scorch Average 94.67
Percentage classification

80 Cottony mold

Ashen mold
60 102
Late scorch 100
40 98
Tiny whiteness
96
20
Normal 94
0 92
Overall average
90
HS H S I HSI 88
Color Features
86
Figure 7: Percentage classification of various diseases 84
82
Early Cottony Ashen Late Tiny Normal
The recognition rate for NN classification strategy of all scorch mold mold scorch whiteness
models was also computed for early scorch, cottony mold and
normal leaf images based upon Algorithm 1, the obtained
results for Model 1 and 5 are reported in Table 3. Figure 8: Classification results per class for neural network
with back propagation.
The convergence curve for the learning step of the neural
Table 3: Recognition rates of individual plant diseases network in the proposed study in this work is better than that of
Model Color Early Cottony Normal Overall [1] as shown in Figure 9. In Figure 9 Approach 1 represents the
Features scorch mold average study presented in [1], while Approach 2 represents our
M1 HS 99 100 100 99.66 proposed study.
M5 HSI 96 98 100 98.00
It can be implied from Table 3 that Model M1 which has used
only the H and S components in computing the texture
features, has emerged as the best model among the various
models. Furthermore, it can be observed from Table 3 that
Model M5 has less recognition rate than Model 1; this was in
part because of the elimination of the intensity (I component)
from computing the texture features in Model M1. As a matter
of fact, elimination of intensity is serviceable in this study
because it nullifies the effect of intensity variations. The
numbers of leaf samples that were classified into each of the
five tested categories using model M1 with specific threshold
value are shown in Table 4 and Figure 8. It is observed from
Table 4 that only few samples from late scorch and tiny
whiteness leaves were misclassified, also, three test images
were misclassified for the case of late scorch infected leaves.
Similarly, in the case of tiny whiteness images, only two test
images from the class were misclassified. In average, accuracy Figure 9: The convergence curve of [1], that is “Approach 1”,
of classification using our approach was 94.67 compared to and our approach “Approach 2”.
92.7 in case of using the approach presented in [1].

36
 International Journal of Computer Applications (0975 – 8887)
Volume 17– No.1, March 2011

It can be concluded from the above tables and figures that the REFERENCES
obtained results achieved an acceptable level of optimal results [1]. Al-Bashish, D., M. Braik and S. Bani-Ahmad, 2011.
which can also add more weighting rate to the proposed study. Detection and classification of leaf diseases using K-
The presented approaches in this paper and in [1] have been means-based segmentation and neural-networks-based
implemented in MATLAB under Windows XP environment. classification. Inform. Technol. J., 10: 267-275. DOI:
All the experiments are carried out on a desktop computer with 10.3923/itj.2011.267.275
Intel (R) Pentium (R) CPU 2.20GHZ 645 MHZ and 256MB of
RAM. The average computation time for the two proposed [2]. Aldrich, B.; Desai, M. (1994) "Application of spatial grey
approaches was computed in seconds for the models M1, M2, level dependence methods to digitized mammograms,"
M3, M4 and M5 as shown in Table 5. The data in Table 5 was Image Analysis and Interpretation, 1994., Proceedings of
obtained for the two approaches using the same neural network the IEEE Southwest Symposium on, vol., no., pp.100-105,
structure and under the same machine. Figure 10 shows that 21-24 Apr 1994. DOI: 10.1109/IAI.1994.336675
our proposed approach has 19% speedup over the approach of
[1]. [3]. Ali, S. A., Sulaiman, N., Mustapha, A. and Mustapha, N.,
(2009). K-means clustering to improve the accuracy of
Table 5: Average computation time. decision tree response classification. Inform. Technol. J.,
Model Approach 1 Approach 2 8: 1256-1262. DOI: 10.3923/itj.2009.1256.1262
M1 456.19 392.98 [4]. Bauer, S. D.; Korc, F., Förstner, W. (2009): Investigation
M2 399.00 359.91 into the classification of diseases of sugar beet leaves
M3 477.91 388.27 using multispectral images. In: E.J. van Henten, D.
Goense and C. Lokhorst: Precision agriculture ‟09.
M4 408.22 347.44 Wageningen Academic Publishers, p. 229-238. URL:
M5 449.12 414.60 http://www.precision-crop-protection.uni-
Average 438.09 380.64 bonn.de/gk_research/project.php?project=3_09
[5]. Camargo, A. and Smith, J. S., (2009). An image-
processing based algorithm to automatically identify plant
20 disease visual symptoms, Biosystems Engineering,
18 Volume 102, Issue 1, January 2009, Pages 9-21, ISSN
Percentage speedup

16 1537-5110, DOI: 10.1016/j.biosystemseng.2008.09.030.

14
12 [6]. Ford, A. and Roberts, A., (2010) Colour Space
10 Conversions, August 11, 1998.
8 http://www.poynton.com/PDFs/coloureq.pdf, viewed on
6 May 2010.
4
2 [7]. Hartigan, J. A.; Wong, M. A. (1979). "Algorithm AS 136:
0 A K-Means Clustering Algorithm". Journal of the Royal
HS H S I HSI Statistical Society, Series C (Applied Statistics) 28 (1):
100–108.
Used model
[8]. Hillnhuetter, C. and A.-K. Mahlein, Early detection and
Figure 10: Speedup of computation times (approach 2 over localisation of sugar beet diseases: new approaches,
approach 1). Gesunde Pfianzen 60 (4) (2008), pp. 143–149.

4. CONCLUSION AND FUTURE WORK [9]. Jun, W. and S. Wang, (2008). Image thresholding using
weighted parzen-window estimation. J. Applied Sci., 8:
In this paper, respectively, the applications of K-means 772-779. DOI: 10.3923/jas.2008.772.779. URL:
clustering and Neural Networks (NNs) have been formulated http://scialert.net/abstract/?doi=jas.2008.772.779
for clustering and classification of diseases that affect on plant
leaves. Recognizing the disease is mainly the purpose of the [10]. MacQueen, J.. (1967). Some methods for
proposed approach. Thus, the proposed Algorithm was tested classification and analysis of multivariate observations. In
on five diseases which influence on the plants; they are: Early L. M. LeCam and J. Neyman, editors, Proceedings of the
scorch, Cottony mold, ashen mold, late scorch, tiny whiteness. Fifth Berkeley Symposium on Mathematical Statistics and
Probability, volume 1, pages 281--297, Berkeley, CA,
The experimental results indicate that the proposed approach is
1967. University of California Press. URL:
a valuable approach, which can significantly support an
http://projecteuclid.org/DPubS?service=UI&version=1.0
accurate detection of leaf diseases in a little computational
&verb=Display&handle=euclid.bsmsp/1200512992.
effort.
An extension of this work will focus on developing hybrid [11]. Martinez, A., (2007). 2007 Georgia Plant Disease
algorithms such as genetic algorithms and NNs in order to Loss Estimates. http://www.caes.uga.edu/Publications/
increase the recognition rate of the final classification process displayHTML.cfm?pk_id=7762. Viewed on Saturday, 15,
underscoring the advantages of hybrid algorithms; also, we will January, 2011.
dedicate our future works on automatically estimating the [12]. Otsu, N. (1979). "A threshold selection method from
severity of the detected disease. gray-level histograms". IEEE Trans. Sys., Man., Cyber. 9:
62–66. DOI:10.1109/TSMC.1979.4310076.

37
 International Journal of Computer Applications (0975 – 8887)
Volume 17– No.1, March 2011

[13]. Otsu, N., "A Threshold Selection Method from Gray- [18]. Rumpf, T., A.-K. Mahlein, U. Steiner, E.-C. Oerke,
Level Histograms," IEEE Transactions on Systems, Man, H.-W. Dehne, L. Plumer, Early detection and
and Cybernetics, Vol. 9, No. 1, 1979, pp. 62-66. classification of plant diseases with Support Vector
Machines based on hyperspectral reflectance, Computers
[14]. Prasad Babu, M. S. and Srinivasa Rao , B. (2010) and Electronics in Agriculture, Volume 74, Issue 1,
Leaves recognition using back-propagation neural October 2010, Pages 91-99, ISSN 0168-1699, DOI:
network - advice for pest and disease control on crops. 10.1016/j.compag.2010.06.009.
Technical report, Department of Computer Science &
Systems Engineering, Andhra University, India. [19]. Wang, X., M. Zhang, J. Zhu and S. Geng, Spectral
Downloaded from www.indiakisan.net on May 2010. prediction of Phytophthora infestans infection on
tomatoes using artificial neural network (ANN),
[15]. Sezgin, M. and Sankur, B. (2003). "Survey over International Journal of Remote Sensing 29 (6) (2008),
image thresholding techniques and quantitative pp. 1693–1706.
performance evaluation". Journal of Electronic Imaging
13 (1): 146–165. DOI:10.1117/1.1631315. [20]. Weizheng, S., Yachun, W., Zhanliang, C., and
Hongda, W. (2008). Grading Method of Leaf Spot
[16]. Soltanizadeh, H. and B.S. Shahriar, 2008. Feature Disease Based on Image Processing. In Proceedings of the
extraction and classification of objects in the rosette 2008 international Conference on Computer Science and
pattern using component analysis and neural network. J. Software Engineering - Volume 06 (December 12 - 14,
Applied Sci., 8: 4088-4096. DOI: 2008). CSSE. IEEE Computer Society, Washington, DC,
10.3923/jas.2008.4088.4096. URL: 491-494. DOI= http://dx.doi.org/10.1109/CSSE.
http://scialert.net/abstract/?doi=jas.2008.4088.4096. 2008.1649.
[17]. Stone, M. C. (August 2001). “A Survey of Color for
Computer Graphics”. Course at SIGGRAPH 2001.

38