International Journal of Science and Engineering Applications

Volume 7–Issue 08,143-147, 2018, ISSN:-2319–7560

Support Vector Machine Based Classification of Leaf

Diseases

Ko Ko Zaw Dr. Zin Ma Ma Myo Daw Thae Hsu Thoung

Department of Electronic Department of Electronic Department of Electronic
Engineering Engineering Engineering
Technological University Technological University Technological University
Thanlyin, Myanmar Thanlyin, Myanmar Thanlyin, Myanmar

Abstract: Myanmar is well known for agricultural country; wherein about 65% of the labor force depends on agriculture. Since the
leaf diseases are microscopic organism, cannot be detected normal human eyes. Leaves are special indicator to distinguish the diseases
because the image information of the leaf are changed when the leaf surf the diseases. So, the image processing techniques can be used
in agricultural sector. The research work presents a support vector machine classifier algorithm by using MATLAB R2017a for the
classification of leaf diseases such as Alternaria Alternata, Cercospora leaf spot, Bacterial Blight and so on. In this research work,
RGB color space is converted into HSI (Hue Saturation Intensity) color space. In segmentation step, k-means clustering is used to
select the defected area, and it is extracted the features by using GLCM (Gray Level Co-occurrence Matrix). Prior to the features
extraction, the median filter is used for getting noise free feature results. Finally, the leaf disease is classified by using support vector
machine (SVM) and computes the accuracy. From the obtained results, the maximum accuracy of the system is 83%.

Keywords: leaf diseases; median filter; k-means clustering; gray level co-occurrence matrix; support vector machine

robust and accurate disease detection system [2]. While

1. INTRODUCTION working with image processing, on the other hands, the
Most of the diseases symptoms are found in leaves, stem and
database collection is the most challenging task. For database
fruit. The image processing can be used in the leaf diseases
collection, it is necessary to collect the basic information
detection and classification system. The common diseases of
about the crop and its diseases as the important task.
leaf are Bacterial Blight, Anthracnose, Alternaria Alternata,
Therefore, a detail study should be done on the types of
and so on. Such diseases are commonly found on mango, rice,
disease, their symptoms on crop and the patterns of disease.
watermelon, and others leaf. Leaf diseases can decrease the
By observing the patterns of disease, the system will get
yield. These diseases may cause by pathogen such as fungi
designed. The mainly occurring diseases on leaf are Bacterial
and bacterial. These diseases can be automatically detected
disease, Fungal disease, Viral disease and diseases due to
and classified by using multiclass support vector machine
insects. The paper gives the detailing of these diseases [3].
(multiclass SVM). The main purpose of this research work is
to design, implement and evaluate an image processing based 3. METHODOLOGY
software solution for classification of leaf diseases. The The methodology of the research work can be divided into
diseases of leaf features are as follow. four stages such as image preprocessing, image segmentation,
1. Alternaria Alternata: small reddish brown circular feature extraction, and disease detection and classification.
spots appear on the leaves. Block diagram of the system is shown in Figure 1 [4].
2. Anthracnose: Appears as small regular or irregular
dull violet or black leaf spots with yellowish halos. Image Preprocessing
Leaves turn yellow and fall out.
3. Bacterial Blight: Appearance of one to several small
water soaked, dark colored irregular spots on leaf. Image Segmentation
4. Cercospora: Leaf spots are minute, brown with
yellow halo. Spots are scattered, circular or irregular
and become dark brown with age [1].
Features Extraction
2. LITERATURE REVIEW
Some papers are describing to detection leaf disease using
various methods suggesting the various implementation ways Disease Detection and Classification
as follow. Visual Analysis, Image Processing and Optical
Sensor are mainly implemented in three ways as the disease
detection method. By using these three methods, the system Figure 1. Block Diagram of the System
can be developed to detect the disease earlier and that can
The flow chart of the leaf diseases detection and classification
overcome the challenges and disadvantages. By means of the
system is shown in Figure 2 [5].
methods comparison, disease detection by using visual
analysis does not give the accurate output while in case of
optical sensor, the system is not easy to implement and costly.
So, image processing is the only way to build the simple,

www.ijsea.com 143
 International Journal of Science and Engineering Applications
Volume 7–Issue 08,143-147, 2018, ISSN:-2319–7560

The procedure of leaf diseases detection and classification

Start system is the following.
1. Read input image.
Read the Image 2. Resize image.
3. The resized image is contrasted enhancement.
4. Converted RGB color space to HSI color space.
5. Apply k-mean clustering operation.
Resize Image, Contrast Enhanced 6. Image is segmented into three sub-features.
Image, HSI Color Space 7. Select the disease affected area.
8. Filter the image by using median filter.
9. Extract the feature from the image using Gray-Level
K - Mean Clustering Co-occurrence Matrix (GLCM).
10. Compute contrast, correlation, energy,
homogeneity, mean standard, entropy, root mean
square.
Image is Segmented into 3 Sub-feature 11. Classify the diseases using multiclass support vector
ROI and Select the Disease Affected ROI machine.
12. Compute the accuracy.
13. Show the accuracy.
Median Filter 3.1 HSI Color Space
HSI model is proposed to improve the RGB model. The Hue
Saturation Intensity (HSI) color model closely resembles the
Extract the Features from Image using GLCM color sensing properties of human vision. The HSI color space
was developed to be ‘intuitive’ in manipulating color and was
designed to approximate the way humans perceive and
Compute Contrast, Correlation, Energy, interpret color. HLS (Hue, Lightness, and Saturation) color
space is similar to the HSI, the term light is rather than
Homogeneity, Mean, Standard, Entropy, RMS
intensity. HSI color space is best for traditional image
processing functions such as convolution, equalization,
histogram and so on, which operate by the manipulation of
Classification and Disease Detection brightness values since I is equally dependent on R,G, and B.
using multiclass SVM The arccos function to compute H, arccos always gives you a
value between 0 and 180 degrees.
However, H can assume values between 0 and 360 degrees. If
Find Disease Types using Database B > G, then H must be greater than 180 degrees. Therefore, if
B > G, just compute H as before and then take (360 degrees –
H) as the actual hue value. The saturation is the distance on
the triangle in the rgb-subspace from white relative to the
Classified the Disease
distance from white to the fully saturated color with the same
hue. Fully saturated colors are on the edges of the triangle.
Hue is a color attribute that describes a pure color (pure
Show the Disease Type yellow, orange or red), whereas a saturation gives a measure
of degree to which a pure color is diluted by white light.

Compute the Accuracy

Show the Accuracy

End

Figure 2. The Flow Chart of Leaf Diseases Classification System

The purposes and the advantages using image analysis are as

follow [6].
1. To identify diseased leaf,
2. To measure affected area by disease,
3. To find the boundaries of the affected area, Figure 3. HSI Color Space
4. To find out the color of the affected area,
5. To identify the diseases correctly.

www.ijsea.com 144
 International Journal of Science and Engineering Applications
Volume 7–Issue 08,143-147, 2018, ISSN:-2319–7560

Table 1. Some Colors and Three HIS Intensity Value

Color H S I
Red 60º 1 0.375
Green 120º 1 0.375
Blue 240º 1 0.375
Yellow 60º 1 0.375
Cyan 180º 1 0.375 (c)
Figure 4. Image Preprocessing (a) Original Image, (b) Resize Image,
Magenta 300º 1 0.375 and (c) Contrast Enhanced Image

White - 0 0.75 Histogram equalization is a consideration for the image

enhancement. It is a traditional approach of image contrast
Black - 0 0 adjustment then the histogram equalization is shown in Figure
5.

The HSI color model owes its usefulness to two principle

facts. Firstly, the intensity component, I, is decoupled from
the color information in the image. Secondly, the hue and
saturation components are intimately to the way in which
human being perceive color. HSI color space is shown in
Figure 3 [7] and HSI intensity of some colors is shown in
Table 1 [8].

4. TEST AND RESULTS

In this section, tests and results of automated leaf diseases
detection and classification system using gray-level co-
occurrence matrix (GLCM) and multiclass support vector
machine (multiclass SVM) are expressed. The performance of
the algorithm as a whole is analyzed and discussed. Tests
were carried out to find the best segmentation result, so that
the error measurement is minimized, and to confirm that leaf
diseases detection and classification can perform accurately.
As were as confirming that the system provides accurate
detection, experiments were also conducted in order to
confirm the detection by extracting the properties of the (a)
images. In the image preprocessing step consist of three parts
such as read the image, resize image and contrast enhanced
image. The image preprocessing is shown in Figure 4.

(a)

(b)
Figure 5. Histogram Equalization Image (a) before contrast
enhancement, (b) after contrast enhancement

After making the preprocessing, the RGB color space is

(b) converted into HSI and L*a*b (a luminosity layer 'L*',

www.ijsea.com 145
 International Journal of Science and Engineering Applications
Volume 7–Issue 08,143-147, 2018, ISSN:-2319–7560

chromaticity-layer 'a*' and 'b*'). The HSI color space is higher In the feature extraction, the segmented image is converted
peak signal to noise ratio than L*a*b color space. The results into gray scale image, then filter by using median filter and
are shown in Figure 6. GLCM is used to extract the feature [9]. The segmented
image is as shown in Figure 8. The gray scale image and
median filter image are shown in Figure 9 and Figure 10.
After that, the affected area of leaf can be calculated. The
affected area of leaf is shown in Figure 11. Finally, the leaf
diseases can be classified by using multiclass support vector
machine. The disease type is shown in Figure 12. The
accuracy is shown in Figure 13.

(a)

Figure 8. Segmented Image

(b)

Figure 9. Gray Scale Image

(c)
Figure 6. Color Space (a) HIS Color Space, (b) L*a*b Color Space,
and (c) Peak Signal to Noise Ratio (PSNR) of Color Space

The image is segmented into three sub-features by using k- Figure 10. Median Filter Image
mean clustering. And then, select the disease affected area.
The results are shown in Figure 7.

Figure 7. k-mean clustering Figure 11. Affected Area of Leaf

www.ijsea.com 146
 International Journal of Science and Engineering Applications
Volume 7–Issue 08,143-147, 2018, ISSN:-2319–7560

like to thank many colleagues from digital image processing

research group of Department of Electronic Engineering of
Technological University (Thanlyin). The author particularly
wishes to acknowledge all the teachers from Department of
Electronic Engineering, Technological University (Thanlyin),
for their support, encouragement and invaluable guidance in
preparation of this research. The authors would like to express
Figure 12. Disease Type of Leaf their thanks to all persons who have given support during the
preparation period of this research work.

7. REFERENCES
[1] Ko Ko Zaw, Zin Ma Ma Myo, and Wah Wah Hlaing.
2018. “Multiclass Support Vector Machine Based
Detection and Classification of Leaf Diseases”, 11th
National Conference on Science and Engineering, YTU,
Yangon, Myanmar
Figure 12. Disease Type of Leaf [2] A. Miyatra, S. Solanki. 2014. “Disease and Nutrient
Deficiency Detection in Cotton Plant,” International
Journal of Engineering Development and Research, vol.
This system elaborates the evaluated results with proper 2, no 2, pp. 2801-2804.
discussion. Windows 10 based system 4GB DDR4 Memory,
1TB of HDD, Core i5 is used for conduction the experiments. [3] V. A. Gulhane, A. A. Gurjar. 2011. “Detection of disease
And then, MATLAB R2017a is used for the simulation of on cotton Leaves and Its Possible Diagnosis”,
work. The total program execution time is shown in Figure International Journal of Image Processing (IJIP), vol. 5 :
14. no 5 : pp. 590-598.
[4] S. S. Panchal, R. Sonar. 2016. “Pomegranate Leaf Disease
Detection Using Support Vector Machine,” International
Journal of Engineering and Computer Science, vol. 5, no
6, pp. 16815-16818.
[5] V. M. Tiwari, T. Gupta. 2017. “Plant Leaf Disease
Analysis using Image Processing Technique with
Figure 14. The Total Program Execution Time Modified SVM-CS Classifier,” International Journal of
Engineering and Management Technology, vol. 5, no. 1,
pp. 11-17.
Accuracy is defined as the ratio of the number of correctly
recognized samples to the total number of test samples [10]. [6] A. N. Rathod , B. A. Tanawala, V. H. Shah. 2014. “Leaf
The percentage values of accuracy of each disease are shown disease detection using image processing and neural
in Table 2. network”, IJAERD vol. 1, no. 6, June, pp. 1-10.
[7] S. Jayaraman, S. Esakkirajan , and T. Veerakumar. 2009.
Digital Image Processing. U.S.A.: McGraw Hill Co,.
Table 2. The percentage values of accuracy of each disease
[8] Myo Myo Han. 2017. “Analysis on Detection
Percentage of Improvement of Textile Fabric Defects using Gray Level
Sr. No Disease Type
Accuracy Co-occurrence Matrix for automatic system,” M.E. thesis,
1. Alternaria Alternata 80.6452 % Dept. Electronic Eng., Yangon Technological
University., Yangon, Myanmar.
2. Anthracnose 82.2581 %
[9] D. Gadkari. 2004. “Image Quality Analysis Using
3. Bacterial Blight 80.6452 % GLCM,” M.E. thesis, University of Central Florida,
4. Cercosporal Leaf Spot 82.2581 % Orlando, Florida.
[10] S. S. Patki, G. S. Sable. 2016. “Cotton Leaf Disease Detection &
5. Healthy Leaf 83.8710 %
Classification using Multi SVM,” International Journal of
Advanced Research in Computer and Communication
Engineering, vol. 5, no. 10, pp. 165-168.
5. CONCLUSION
From the results shown in above, the total program execution
time takes 0.24045 minute and the maximum accuracy is
83.8710%. Furthermore, this system can be applied for the
other leaf diseases to get the peak signal to noise ratio. When
the total program execution time is less, the system can be
assigned the higher performance.

6. ACKNOWLEDGEMENTS
Firstly, the author would like to acknowledge Dr. Thein Gi,
Rector of Technological University (Thanlyin), for her kind
permission to carry out this research work. The authors would

www.ijsea.com 147