CHAPTER 9

IDENTIFICATION OF LEAF DISEASE

FROM DIGITAL MULTISAMPLING
AND CLASSIFICATION USING SUPPORT
VECTOR MACHINE CLASSIFIER

VANDITA ANAND, ANJANA PANDEY

Abstract
Plant disease contributes significantly to a deterioration in the quantity and
quality of agricultural goods. Studies on plant diseases often involve
examining visually observable patterns in plants. Image processing
methods can be employed to identify and analyze plant diseases. Diseased
symptoms are frequently visible on the leaves, stems, and fruit. This
method is composed of five steps: (i) collection of images, (ii) image pre-
processing, (iii) segmentation, (iv) extraction of features, and (v)
classification. The image of the plant leaf is divided into segments, and
features are calculated using clusters. The disease influences the clusters
using the K-means clustering technique and a multi-support vector
machine (SVM) algorithm for identification. The experimental result
showed that the suggested method could perfectly detect and classify the
four prevalent diseases on plant leaves: Alternaria alternata, Anthracnose,
Bacterial blight, and Cercospora leaf spot. The efficiency of leaf disease
detection was observed at 97.23% via a support vector machine (SVM)
classifier.

Keywords: Image Processing; Image Segmentation; K-Means clustering

technique; Leaf Disease Detection; Support Vector Machine (SVM)
126 Chapter 9

Graphical abstract

Introduction
India is a predominantly agricultural country, with agriculture as the
economic backbone. This is because nature has endowed it with vast tracts
of fertile land and an adequate water supply, essential components of an
agro-based economy. Agriculture has major problems with diseases caused
by pests and insects. Accurate diagnosis and immediate intervention are
essential to prevent serious losses of crops. Diseases may affect many
plant parts, such as the fruit, stem, and leaves. Plant diseases frequently
cause disease epidemics that cause starvation and massive fatalities
(Arivazhagan et al., 2013).

A natural occurrence, diseases can significantly impact plants by lowering

growth, traits, and production. The leaf is the pivotal portion of the plant
to evaluate for plant diseases (Tiwari et al., 2023a). Identifying and
classifying leaf diseases effectively is essential for avoiding agricultural
losses. Different plant leaves carry different diseases. Plant leaf diseases
are divided into three categories: viral, fungal, and bacterial (Weizheng et
al., 2008). Alternaria alternata, Anthracnose, Bacterial blight, and Cercospora
leaf spot are the most frequent plant diseases.

Matlab is the most used tool for developing image processing and SVM
classifier algorithms. Matlab's built-in image processing features benefit
from utilizing it as an image processing method development (Rojas,
 Identification of Leaf Disease from Digital Multisampling and 127
Classification Using Support Vector Machine Classifier

1996; Tiwari et al., 2023b). Plant diseases create significant output and
economic losses in the country. Human eyes cannot identify specific
diseases, and the symptoms are not evident. Various methods for detecting
diseases on plant leaves, including image processing approaches. Using
photos based on the k-means clustering algorithm to identify leaf diseases
was investigated (Barbedo, 2013). For leaf disease detection and diagnosis,
pictures of a leaf are identified using a multiclass SVM. Various forms of
bacterial, fungal, and viral leaf spot diseases in agriculture were researched
for automatic leaf disease identification utilizing digital image processing
(Rathod et al., 2013). Feature extraction is utilized to identify properties
such as border, form, color, and texture to recognize leaf disease (Patil et
al., 2014). The SVM technique is utilized to classify crops and weeds in
digital pictures. According to the classification rate, SVM delivers 97%
accuracy of crops as weeds (Ahmed et al., 2012).

The primary focus of this research was to employ image processing

techniques and an SVM classifier to investigate various forms of diseases
seen in plant leaves. In this article, the five primary processes of the
automated plant leaf disease detection are capture, image pre-processing,
segmentation, feature extraction, and classification. Images of diseased
leaves are taken and saved for future studies. After that, photos are
subjected to image-enhancing preparation. The K-Means clustering
algorithm segments the captured leaf pictures and creates clusters. SVM
was used to classify and identify plant leaf diseases such as Alternaria
alternata, Anthracnose, Bacterial blight, and Cercospora leaf spot using
GLCM characteristics extracted using K-Means.

Literature review
This section discusses several image processing techniques for detecting
plant diseases. Depending on the applications, many solutions have been
developed to address or at least minimize the issues, using image processing
and certain automated classification techniques.

The method to categorize the watermelon leaf diseases anthracnose and

downy mildew was proposed by Kutty et al. (2013). Based on the RGB
color component, this region of interest has to be distinguished from the
infected leaf sample. The median filter is then used to segment data and
minimize noise. Additionally, the toolbox for neural network pattern
recognition is employed for categorization. Based on its RGB mean color
component, the proposed technique has an accuracy rate of 75.9%.
128 Chapter 9

The objective of Sannaki et al. (2013) is to use image processing and

artificial intelligence techniques on photographs of grape plant leaves to
identify the illness. They mostly categorize grape leaf downy mildew and
grape leaf powdery mildew. Anisotropic Diffusion is employed to
preserve the information on the impacted area of the leaf. Feed Forward
Back Propagation Network classifier is used for classification. They just
utilized the Hue function, which produces more accurate results.

Akhtar et al. (2013) used the SVM technique to classify and detect rose
leaf diseases such as black spots and anthracnose. Authors have employed
the threshold approach for segmentation, and the threshold values have
been set using Ostu's algorithm. Eleven haralick features based on DWT,
DCT, and texture are extracted in this method, which is then combined
with SVM to get accurate results.

Scab, apple rot, and apple blotch were the three apple fruit diseases
examined by Dubey et al. (2012) using K-means clustering for segmentation.
The segmented picture is then used to extract features. Multiclass SVM is
the technique used for classification.

According to Mokhtar et al. (2015), Powdery mildew and Early blight are
the two diseases that affect tomato leaves. Image improvement methods
include smoothing, reducing noise, enlarging the image, isolating the
image, and removing the backdrop. The Gabor wavelet transformation is
used to extract features from feature vectors and to classify data. SVM
employs the Cauchy, Laplacian, and Invmult kernels for output assessment
and training in disease detection.

Khirade et al. (2015) highlighted the core processes involved in classifying

and detecting plant diseases using images, i.e., image capture, image pre-
processing, image segmentation, feature extraction, and classification are
some of the procedures involved. Artificial neural networks and back
propagation neural networks are two classification techniques used. Oo et
al. (2018) demonstrate that the proposed system can successfully detect
and classify four major plant leaves diseases: Bacterial blight, Cercospora
leaf spot, Powdery mildew and Rust.

The categorization of cotton leaf disease analysis by Bhong et al. (2016)

has adopted the neural network approach. K-means clustering has been
used for segmentation. Different cotton leaf diseases include Alternaria,
Cercospora, Red spot, White spot, and Yellow spot on the leaf. The
MATLAB toolbox (18B) has been used for experiments.
 Identification of Leaf Disease from Digital Multisampling and 129
Classification Using Support Vector Machine Classifier

Gavhale et al. (2014) described various image processing methods to

remove the unhealthy portion of leaves. DCT domain image enhancement
and color space conversion are done during pre-processing. Segmentation
then occurs using the k-means clustering approach. Use of the GLCM
Matrix is made for feature extraction. SVM with radial basis kernel and
polynomial kernel is used to classify citrus leaf canker and anthracnose
disease.

Materials and methods

The steps for integrating image processing to identify and categorize
diseased leaf as follows in Fig 9.1.

Leaf image acquisition RGB (Red, Green and Blue) formatted image

Remove noise in the image, Image smoothing,

Leaf image Pre-processing Image enhancement, RGB images into the grey
images, and Histogram equalization

Otsu’s method, k-means clustering, converting

Image Segmentation RGB image into HIS model

Color, texture, morphology, etc. are the

Feature Extraction characteristics, GLCM, LBP

Classification is done by using Support Vector

Detection and Classification of Machine (SVM).
leaf disease

Figure 9.1: Basic processes for detecting and identifying plant diseases are
depicted

Image acquisition
The process of image acquisition entails using a camera to capture
pictures. Among the 14 picture samples were captured of five classes:
Alternaria alternata, Anthracnose, Bacterial blight, Cercospora leaf spot
and Healthy leaf. Our investigation focused on the diseased photographs
130 Chapter 9

of plant leaves that were saved on digital media for later MATLAB
operations. Images of a sample diseased leaf are shown in Fig. 9.2.

Image pre-processing
Image pre-processing was used to eliminate noise from a picture or
remove other objects. Various pre-processing approaches were investigated.
The image was smoothed using the smoothing filter. Image enhancement
is done to boost contrast.

Using the color conversion equation, convert RGB images to greyscale

images (1)

f(x)=0.2989*R + 0.5870*G + 0.114.*B - - - - - - - - - - - - - (1)

The histogram equalization was performed to enhance the image of

diseased plant photographs by dispersing the image intensities. The
cumulative distribution function was used to disperse the intensity values
(Thangadurai et al., 2014).

Image segmentation
The term "segmentation" refers to separating an image into segments with
comparable qualities. Several techniques can be used to segment the data,
including the Otsu method, k-means clustering, and converting RGB
images to HIS models (Gonzalez et al., 2007). Picture segmentation aids
in object recognition, spot recognition and image border edge detection to
locate the contaminated portion of the leaf. The 8 connectivity of pixels is
considered while determining the border, and the boundary detection
technique is used. Objects were divided into K classes using the K-means
clustering technique based on characteristics. The sum of the squares of
the distance between the object and the relevant cluster was used to
classify it (Al-Hiary et al., 2011).

Otsu Threshold Algorithm:

Thresholding converts grayscale pictures into binary by setting all pixels

below a certain threshold to zero and all pixels above that threshold to one
(Padmavathi et al., 2016).

The color of the diseased leaf changes, indicating the disease's symptoms.
As a result, the color of the leaves may be utilized to identify the diseased
region of the leaf, as shown in Fig 9.2. The R, G, and B components of the
 Identification of Leaf Disease from Digital Multisampling and 131
Classification Using Support Vector Machine Classifier

picture were separated. The Otsu technique is used to determine the

threshold. If the green pixels' intensity fell below a certain level, they were
concealed and erased (Badnakhe et al., 2012).

Figure 9.2: Otsu Segmentation

Feature extraction
Feature extraction is necessary for object recognition. Feature extraction is
utilized in various image processing applications. Plant disease may be
identified using color, texture, morphology, edges, and other features
(Khirade et al., 2015).

A statistical technique is Grey Level Co-occurrence Matrices. An essential

classification feature extraction approach computes the connection
between pixel pairs in the picture. To extract disease symptoms, calculate
the Skewness, Standard Deviation, Homogeneity, Contrast, Smoothness,
Correlation, Kurtosis, Energy, Entropy, Mean, Variance, RMS, and IDM
values for each GLCM texture feature (Al-Hiary et al., 2011).
132 Chapter 9

Classification
The categorization uses an SVM once the color and texture data have been
extracted. In a high- or infinite-dimensional space, an SVM creates a
hyper-plane or set of hyper-planes, which can be used for categorization,
analysis, or other purposes. For data analysis and pattern identification in
classification and regression analysis, a supervised learning model with
concomitant learning techniques is called an SVM (Weizheng et al.,
2008).

The training and validation methods were among the most important
components of utilizing SVM to create a detailed process model. The
training and evaluations of datasets were split into two parts: a training
feature set for training the SVM model and a testing feature set for
validating the trained SVM model's accuracy. Finally, comparing disease
data to leaf data, the percentage of disease-affected areas is computed for
the disease category with the best accuracy value (Oo et al., 2018). A
proposed schematic flow chart is shown in Fig. 9.3.

Figure 9.3: Proposed Flow Chart

Results and discussions

Image processing of leaf images was performed using MATLAB 2018b
software. The k-means clustering approach was used to extract the
characteristics from the disease-affected clusters. Fourteen samples were
taken for diseases, i.e., Alternaria alternata, Anthracnose, Bacterial blight,
Cercospora leaf spot and healthy leaf.
 Identification of Leaf Disease from Digital Multisampling and 133
Classification Using Support Vector Machine Classifier

Image acquisition
Images of the healthy and diseased leaves were imported from the
database, as shown in Fig. 9.4. This image is uploaded into MATLAB and
analyzed.

Sample 1 Sample 2

Sample 3 Sample 4

Sample 5 Sample 6

Sample 7 Sample 8
134 Chapter 9

Sample 9 Sample 10

Sample 11 Sample 12

Sample 13 Sample 14
Figure 9.4: Sample diseased leaf images

Image pre-processing
Figure 9.5 depicts image pre-processing, converting RGB images to
greyscale images via color conversion, which is pivotal in transforming
RGB images into grayscale representations, retaining essential luminance
information while discarding chrominance components. Watermelon leaf
diseases Anthracnose and downy mildew were categorized using this
method. The region of interest from an infected leaf sample must be
identified using the RGB color component (Kutty et al., 2013).
 Identification of Leaf Disease from Digital Multisampling and 135
Classification Using Support Vector Machine Classifier

Figure 9.5: Image Pre-processing of diseased leaf sample

Image segmentation
The clusters of the unhealthy leaf are shown in Fig. 9.6, with cluster 3
representing the diseased portion and other clusters representing the
healthy parts. Image processing techniques were applied to grape plant
leaves to predict diseases, focusing on downy mildew and powdery
mildew (Sannaki et al., 2013). For validation, masking is used to remove
the backdrop. The data is split using the k-means clustering technique.
Following segmentation, features have been extracted using the Gray
Level Co-occurrence Matrix (Gavhale et al., 2014).

Figure 9.6: Image segmentation of diseased leaf sample

136 Chapter 9

Feature extraction & classification

Out of fourteen leaf samples, 06 were categorized as Alternaria
alternata disease, 03 as Anthracnose disease, 02 as Bacterial blight disease,
01 as Cercospora leaf spot disease, and 02 as healthy leaves. The accuracy
for four diseases, Alternaria alternata, Anthracnose, Bacterial blight and
Cercospora leaf spot, was 98.38%, 98.38%, 98.38% and 96.77%, respectively,
with average classification accuracy is 97.23%. Table 9.1 describes the
characteristics of Contrast, Correlation, Energy, Homogeneity, Mean,
Standard Deviation and Variance. These characteristics were used to
identify and categorize the leaf disease (Fig. 9.7).

Figure 9.7: Feature extraction of diseased leaf sample

Table 9.1: Accuracy of SAMPLES of plant leaves

Sample No. Disease name Affected region Accuracy %

%
1 Alternaria alternata 38.10 % 96.77 %
2 Alternaria alternata 51.36 % 98.38 %
3 Anthracnose 24.62 % 96.77 %
4 Alternaria alternata 82.80 % 96.77 %
5 Bacterial blight 61.57 % 98.38 %
6 Alternaria alternata 60.78 % 98.38 %
7 Bacterial blight 15.00 % 96.77 %
8 Alternaria alternata 45.77 % 95.16 %
9 Anthracnose 15.00 % 98.38 %
10 Alternaria alternata 16.34 % 98.38 %
11 Healthy leaf None 96.77 %
12 Anthracnose 15.06 % 98.38 %
13 Cercospora leaf spot 74.79 % 96.77 %
14 Healthy leaf None 96.77 %
 Identification of Leaf Disease from Digital Multisampling and 137
Classification Using Support Vector Machine Classifier

Conclusions
This study uses an SVM classifier to detect leaf disease, employing K-
means clustering for segmentation. Features such as contrast, correlation,
energy, homogeneity, mean, standard deviation, and variance were
retrieved for infected plants. The diseased leaves considered during the
simulation were Alternaria alternata, Anthracnose, Bacterial Blight and
Cercospora Leaf Spot. Clusters 1 and 3, associated with diseases, were
used to compute features. The classifier used the attributes to identify and
classify the diseases. Out of fourteen leaf samples, 06 were categorized as
Alternaria alternata disease, 03 as Anthracnose disease, 02 as Bacterial
Blight disease, 01 as Cercospora Leaf Spot disease, and 02 as healthy
leaves. Alternaria alternata, Anthracnose, Bacterial blight, and Cercospora
leaf spot had classification accuracy of 98.38%, 98.38%, 98.38% and
96.77%, respectively, with an overall classification accuracy of 97.23%.

Declaration of competing interest

All authors declare no competing interest.

Acknowledgement
The authors are thankful to the Director of MNNIT & Department of
Biotechnology MNNIT Allahabad, Prayagraj, India.

References
Ahmed F., Al-Mamun H.A., Bari A.S.M.H., Hossain E., Kwan P. (2012).
Classification of crops and weeds from digital image: A support vector
machine approach. Crop detection, 40.
Akhtar, Asma, Khanum A., Khan S.A, Shaukat A. (2013). Automated
Plant Disease Analysis (APDA): Performance Comparison of Machine
Learning Techniques. IEEE International Conference on Frontiers of
Information Technology (FIT).
Al-Hiary H., Bani-Ahmad S., Reyalat M., Braik M, AlRahamneh Z.
(2011). Fast and accurate detection and classification of plant diseases.
International Journal of Computer Applications,17(1).
Arivazhagan S., Newlin Shebia R. (2013). Detection of unhealthy region
of plant leaves and classification of plant leaf diseases using texture
features. Agricultural Engineering Institute: CIGR journal, 15(1).
Badnakhe M.R., & Deshmukh P.R. (2012). Infected Leaf Analysis and
Comparison by Otsu Threshold and k-Means Clustering. International
138 Chapter 9

Journal of Advanced Research in Computer Science and Software

Engineering, 2(3).
Barbedo J.G.A. (2013). Digital image processing techniques for detecting,
quantifying and classification plant disease. SpringerPlus, 2(1).
Bhong, V.S., Pawar B.V. (2016). Study and Analysis of Cotton Leaf
Disease Detection Using Image Processing. International Journal of
Advanced Research in Science, Engineering and Technology.
Gavhale K.R., Gawande U., Hajari K.O. (2014). Unhealthy Region of
Citrus Leaf Detection Using Image Processing Techniques. IEEE
International Conference for Convergence of Technology.
Gonzalez R.C. and Woods R.E. (2007). Digital Image Processing, Pearson
Education, Third Edition.
Jalal A.S., Dubey S.R. (2012). Detection and Classification of Apple Fruit
Diseases Using Complete Local Binary Patterns. IEEE Third International
Conference on Computer and Communication Technology.
Khirade S.D., Patil A.B. (2015). Plant Disease Detection Using Image
Processing. Int Conf on Comp Communication Control and
Automation IEEE, 153.
Kutty S.B., Abdullah N.E., Hashim H., Rahim A.A., Kusim A.S., Yaakub
T.N.T., Yunus P.N.A.M., Rahman M.F.A. (2013). Classification of
Watermelon Leaf Diseases Using Neural Network Analysis. IEEE,
Business Engineering and Industrial Applications Colloquium
(BEIAC).
Mokhtar U., Alit M.A.S., Hassenian A.E., Hefny H. (2015). Tomato
leaves diseases detection approach based on support vector machines.
IEEE.
Oo Y.M., Htun N.C. (2018). Plant Leaf Disease Detection and
Classification using Image Processing. International Journal of
Research and Engineering, 5(9).
Padmavathi K., Thangadurai K. (2016). Implementation of RGB and Gray
scale images in plant leaves disease detection –comparative study.
Indian J. of Sci. and Tech., 9.
Patil S.P., Zambre R.S. (2014). Classification of Cotton Leaf Spot Disease
using Support Vector Machine. Int. Journal of Engineering Research
and Applications, 4.
Rathod A.N., Tanawal B., Shah V. (2013). Image Processing Techniques
for Detection of Leaf Disease. 3(11).
Rojas R. (1996). Neural Networks: A Systematic Introduction
Sannaki S.S., Rajpurohit V.S., Nargund V.B., Kulkarni P. (2013).
Diagnosis and Classification of Grape Leaf Diseases using Neural
Network. IEEE.
 Identification of Leaf Disease from Digital Multisampling and 139
Classification Using Support Vector Machine Classifier

Thangadurai K., Padmavathi K. (2014). Computer Vision Image

Enhancement for Plant Leaves Disease Detection. World Congress on
Comp and Communication Tech.
Tiwari N., Gupta B.K., Prakash A., Husain M., & Singh D. (2023a).
Applying Deep Learning and Machine Learning Algorithms for The
Identification of Medicinal Plant Leaves Based on Their Spectral
Characteristics. Journal of Advanced Zoology, 44(S5), 1080–1090.
[Link]
Tiwari N., Gupta B.K., Prakash A., Tiwari K., Alshmrany S., Ali A.,
Husain M., & Singh D. (2023b). Developing A Neural Network-Based
Model for Identifying Medicinal Plant Leaves Using Image
Recognition Techniques. Journal of Advanced Zoology, 44(S5), 1944–
1958. [Link]
Weizheng S., Yachun W., Zhanliang C., and Hongda W. (2008). Grading
Method of Leaf Spot Disease Based on Image Processing. In
Proceedings of the 2008 international Conference on Computer
Science and Software Engineering (CSSE). Washington, DC, pp. 491-
494.