World Journal of Science and Technology 2012, 2(4):190-194

ISSN: 2231 – 2587

Available Online: www.worldjournalofscience.com

_________________________________________________________________
Proceedings of "Conference on Advances in Communication and Computing (NCACC'12)”
Held at R.C.Patel Institute of Technology, Shirpur, Dist. Dhule,Maharastra,India.
April 21, 2012

Crop disease detection using image segmentation

Tushar H Jaware, Ravindra D Badgujar and Prashant G Patil

Asst. Prof, Dept of E&TC, R C Patel Institute of Technology, Shirpur (MS) India

Abstract
The K-Means clustering technique is a well-known approach that has been applied to solve low-level image segmentation
tasks. This clustering algorithm is convergent and its aim is to optimize the partitioning decisions based on a user-defined
initial set of clusters that is updated after each iteration. In the first step we identify the mostly green colored pixels. Next,
these pixels are masked based on specific threshold values that are computed using Otsu's method, then those mostly green
pixels are masked. The other additional step is that the pixels with zeros red, green and blue values and the pixels on the
boundaries of the infected cluster (object) were completely removed. The experimental results demonstrate that the proposed
technique is a robust technique for the detection of plant leaves diseases.

Keywords: Image Segmentation, K-Means clustering, Crop diseases.

INTRODUCTION
tremendous amount of work but also suffers from two major issues:
India is an agricultural country; wherein about 70% of the excessive processing time and subjectiveness rising from different
population depends on agriculture. Farmers have wide range of individuals. Hence to conduct high throughput experiments, plant
diversity to select suitable Fruit and Vegetable crops. However, the biologist need efficient computer software to automatically extract
cultivation of these crops for optimum yield and quality produce is and analyze significant content. Here image processing plays
highly technical. It can be improved by the aid of technological important role.
support. The management of perennial fruit crops requires close It can be improved by the aid of technological support. The
monitoring especially for the management of diseases that can affect management of perennial fruit crops requires close monitoring
production significantly and subsequently the post-harvest life. especially for the management of diseases that can affect production
In case of plant the disease is defined as any impairment of significantly and subsequently the post-harvest life. In case of plant
normal physiological function of plants, producing characteristic the disease is defined as any impairment of normal physiological
symptoms. A symptom is a phenomenon accompanying something function of plants, producing characteristic symptoms. A symptom is
and is regarded as evidence of its existence. Disease is caused by a phenomenon accompanying something and is regarded as
pathogen which is any agent causing disease. In most of the cases evidence of its existence.
pests or diseases are seen on the leaves or stems of the plant.
Therefore identification of plants, leaves, stems and finding out the METHODOLOGY
pest or diseases, percentage of the pest or disease incidence,
symptoms of the pest or disease attack, plays a key role in The K-Means clustering algorithm is proposed by Mac Queen
successful cultivation of crops. in 1967 which is a partition-based cluster analysis method. It is used
In biological science, sometimes thousands of images are widely in cluster analysis for that the K-means algorithm has higher
generated in a single experiment. These images can be required for efficiency and scalability and converges fast when dealing with large
further studies like classifying lesion, scoring quantitative traits, data sets. However it also has many deficiencies: the number of
calculating area eaten by insects, etc. Almost all of these tasks are clusters K needs to be initialized, the initial cluster centers are
processed manually or with distinct software packages. It is not only arbitrarily selected, and the algorithm is influenced by the noise
points. In view of the shortcomings of the traditional K-Means
Received:; Revised Accepted. clustering algorithm, this paper presents an improved K-means
algorithm using noise data filter. The algorithm developed density-
*Corresponding Author
based detection methods based on characteristics of noise data
Tushar H Jaware where the discovery and processing steps of the noise data are
Asst. Prof, Dept of E&TC, R C Patel Institute of Technology, Shirpur (MS) India added to the original algorithm. By preprocessing the data to exclude
these noise data before clustering data set the cluster cohesion of
Email: : tusharjaware@gmail.com the clustering results is improved significantly and the impact of
noise data on K-means algorithm is decreased effectively and the
World Journal of Science and Technology 2012, 2(4):190-194 191

clustering results are more accurate.

1. RGB image acquisition.
2. Create the colour transformation structure.
3. Convert the colour values in RGB to the space specified in the
colour transformation structure.
4. Apply K-means clustering.
5. Masking green-pixels.
6. Remove the masked cells inside the boundaries of the infected
clusters.
7. Convert the infected (cluster / clusters) from RGB to HSI
Translation.
Fig 2 . Image acquisition and Classification Flow chart
8. SGDM Matrix Generation for H and S.
9. Calling the GLCM function to calculate the features. In details, in step 2 a colour transformation structure for the
RGB leaf image is created, and then, a device-independent colour
10. Texture Statistics Computation. space transformation for the colour transformation structure is
applied in step 3. Steps 2 and 3 are inevitable for carrying out step 4.
START
In this step the images at hand are segmented using the K-Means
clustering technique. These four steps constitute phase 1 whereas,
the infected object (s) is/are determined. In step 5, we identify the
Original images
Training set mostly green colored pixels.
After that, based on specified and varying threshold value that
is computed for these pixels using Otsu's method, these mostly
Read the image
green pixels are masked as follows: if the green component of pixel
intensities is less than the pre-computed threshold value, the red,
Perform feature
extraction 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
Perform the algorithm
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
If and the pixels on the boundaries of the infected cluster (object) were
image
contains completely removed. Steps 5 and 6 form phase 2, and this phase is
details helpful as it gives more accurate disease classification and
identification results with satisfied performance and the overall
X is a disease X is not diseases computation time should become significantly less. The observations
behind steps 5 and 6 were experimentally validated.
Next, in step 7 the infected cluster was then converted from
Store X
RGB format to HSI format. In the next step, the SGDM matrices were
then generated for each pixel map of the image for only H and S
images. The SGDM is a measure of the probability that a given pixel
STOP
at one particular gray-level will occur at a distinct distance and
orientation angle from another pixel, given that pixel has a second
Fig1: Algorithm 1- Basic steps describing the proposed algorithm
particular gray-level. From the SGDM matrices, the texture statistics
for each image were generated.
The proposed approach step - by - step of the image
Concisely, the features set were computed only to pixels
segmentation and recognition processes is illustrated in Algorithm 1.
inside the boundary of the infected areas of the leaf. In other words,
In the initial step, the RGB images of all the leaf samples were
healthy areas inside the infected areas were also removed. Steps 7
picked up. Some real samples of those diseases are shown in Figure
– 10 form phase 3 in which the texture features for the segmented
2. It is obvious from Figure 2 that leaves belonging to early scorch,
infected objects in this phase are calculated. Finally, the recognition
cottony mold, ashen mold and late scorch have significant
process in the fourth phase was performed to the extracted features
differences form greasy spot leaves in terms of colour and texture.
through a pre-trained neural network. For each image in the data set
Also, Figure shows two images; the left image is infected with tiny
the subsequent steps in Algorithm 1 were repeated.
whiteness disease, and the right image is a normal image. However,
The image data of the leaves selected for this study would be
the leaves related to these six classes (early scorch, cottony mold,
collected. Algorithms based on image processing techniques for
ashen mold, late scorch, tiny whiteness and normal) had very small
feature extraction and classification would be designed. Manual
differences as discernible to the human eye, which may justify the
feeding of the datasets, in the form of digitized RGB color
misclassifications based on naked eye.
192 Jaware et al.,

photographs would be done for feature extraction and training the algorithm is O (nkt), where n is the number of objects, k is the
SAS statistical classifier.After training the SAS classifier, the number of clusters, and t is the number of iterations.
test data sets would be used to analyze the performance of accurate 2. It works well when the clusters are compact clouds that are
classification. The whole procedure of analysis would be replicated rather well separated from one another.
for three alternate classification approaches to include; statistical 3. The algorithm is not only simple, but also the results are easily
classifier using the Mahalanobis minimum distance method, neural understandable and it can be easily modelled to deal with
network based classifier using the back propagation algorithm and streaming data. Fig. 3 shows that the k-means algorithm is
neural network based classifier using radial basis functions. simple using the data mining tool KNIME especially there are
Comparison of the results obtained from the three approaches would no missing values in the data set.
be completed and the best approach for the problem at hand would 4. Continual improvements and generalizations of the algorithm
be determined. have ensured its continued relevance and gradually increased
its effectiveness as well.
K-means clustering algorithm
RESULTS
K-means clustering algorithm is simply described as follows:
Input: N objects to be cluster (x1, x2, xn), the number of clusters k;
Output: k clusters and the sum of dissimilarity between each object
and its nearest cluster centre is the smallest;
1. Arbitrarily select k objects as initial cluster centres (m1, m2 ...
mk);
2. Calculate the distance between each object Xi and each cluster
centre, and then assign each object to the nearest cluster,
formula for calculating distance as:

d (Xi, mJ) is the distance between data i and cluster j;

3. Calculate the mean of objects in each cluster as the new cluster Fig 4. Taking infected image as input
centres,

Ni is the number of samples of current cluster i;

4. Repeat 2 & 3 until the criterion function E converged, return
(m),m2 . . . mk).

Fig 5. Selected Crop section.

Fig 3 . Flow chart for K Means Clustering

Advantages of K-Means Clustering

1. This algorithm is relatively scalable and efficient in processing
large data sets because the computational complexity of the
Fig 6. Clipping section of diseased leaf.
World Journal of Science and Technology 2012, 2(4):190-194 193

Fig.10. Normal leaf inage after filtering.

Fig 11. Image of normal leaf after segmenatation.

Fig 7 Filtering of diseased leaf. CONCLUSION

We present a general k-means-based clustering algorithm

that can identify natural clusters in datasets, whether they are
embedded in the original space or subspaces. Like traditional k-
means clustering algorithm, the time complexity of the algorithm is
linear with the number of the data points, the dimensionality of the
data, and the number of clusters in the dataset. The experiment
results show that our algorithm is an efficient algorithm with high
clustering accuracy. Clustering analysis method is one of the main
analytical methods in data mining; the method of clustering algorithm
will influence the clustering results directly.

FUTURE SCOPE

Fig 8. Segmented output & infected part of leaf is detected. We present a general k-means-based clustering algorithm
that can identify natural clusters in datasets, whether they are
embedded in the original space or subspaces. Like traditional k-
means clustering algorithm, the time complexity of the algorithm is
linear with the number of the data points, the dimensionality of the
data, and the number of clusters in the dataset. The experiment
results show that our algorithm is an efficient algorithm with high
clustering accuracy.
Standard versions of k-means algorithms seem be better in
finding high fitness solutions. In the same time results obtained in
standard and genetic versions of k-means algorithms relative to
validity indices are also comparable. During extensive search of
solution space, genetic versions of k-means algorithms most often
find solutions with slightly worse fitness values but at the same time
with exceptionally good values of individual validity indices. Further
Fig 9. Reading image of normal leaf. investigation into this matter could present starting point into
improvement of k-means based image clustering techniques.

REFERENCES

[1] Digital image processing Using MATLAB codes by Dhananjay

Theckedath Tech-Max Publication.
[2] Ali, S. A., Sulaiman, N., Mustapha, A. and Mustapha, N., (2009).
K-means clustering to improve the accuracy of decision tree
response classification. Inform. Technol. J., 8: 1256-1262.
DOI: 10.3923/itj.2009.1256.1262
[3] Hillnhuetter, C. and A.-K. Mahlein, Early detection and
localisation of sugar beet diseases: new approaches, Gesunde
Pfianzen 60 (4) (2008)
194 Jaware et al.,

of Tobacco Leaves, 2008 Congress on Image and Signal

[4] Otsu, N. (1979). "A threshold selection method from gray-level
Processing, IEEE computer society.
histograms". IEEE Trans. Sys., Man., Cyber. 9: 62–66.
DOI:10.1109. [7] The MathWorks, Inc., The Student Edition of MATLAB Version 4
User's Guide, Prentice Hall, ISBN 0-13-184979-4, 1995.
[5] Haykin, S., 1999. Neural Networks: A comprehensive foundation,
Second Edition, Prentice Hall. [8] The MathWorks, Inc., The Student Edition of SIMULINK User's
Guide, Prentice Hall, ISBN 0-13-452435-7, 1995.
[6] Xinhong Zhang & Fan Zhang[2008] Images Features Extraction